2008-07-30 00:29:19 +07:00
|
|
|
/*
|
|
|
|
* xsave/xrstor support.
|
|
|
|
*
|
|
|
|
* Author: Suresh Siddha <suresh.b.siddha@intel.com>
|
|
|
|
*/
|
2012-05-22 09:50:07 +07:00
|
|
|
|
|
|
|
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
|
|
|
|
|
2008-07-30 00:29:19 +07:00
|
|
|
#include <linux/bootmem.h>
|
|
|
|
#include <linux/compat.h>
|
2014-05-30 01:12:43 +07:00
|
|
|
#include <linux/cpu.h>
|
2008-07-30 00:29:19 +07:00
|
|
|
#include <asm/i387.h>
|
2012-02-22 04:19:22 +07:00
|
|
|
#include <asm/fpu-internal.h>
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
#include <asm/sigframe.h>
|
2014-10-25 05:58:07 +07:00
|
|
|
#include <asm/tlbflush.h>
|
2008-07-30 07:23:16 +07:00
|
|
|
#include <asm/xcr.h>
|
2008-07-30 00:29:19 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Supported feature mask by the CPU and the kernel.
|
|
|
|
*/
|
2008-07-30 07:23:16 +07:00
|
|
|
u64 pcntxt_mask;
|
2008-07-30 00:29:19 +07:00
|
|
|
|
2010-07-22 00:03:55 +07:00
|
|
|
/*
|
|
|
|
* Represents init state for the supported extended state.
|
|
|
|
*/
|
2012-08-25 04:13:02 +07:00
|
|
|
struct xsave_struct *init_xstate_buf;
|
2010-07-22 00:03:55 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
static struct _fpx_sw_bytes fx_sw_reserved, fx_sw_reserved_ia32;
|
2014-05-30 01:12:43 +07:00
|
|
|
static unsigned int *xstate_offsets, *xstate_sizes;
|
2014-05-31 04:59:24 +07:00
|
|
|
static unsigned int xstate_comp_offsets[sizeof(pcntxt_mask)*8];
|
2014-05-30 01:12:43 +07:00
|
|
|
static unsigned int xstate_features;
|
2010-07-20 06:05:48 +07:00
|
|
|
|
2010-07-20 06:05:49 +07:00
|
|
|
/*
|
|
|
|
* If a processor implementation discern that a processor state component is
|
|
|
|
* in its initialized state it may modify the corresponding bit in the
|
|
|
|
* xsave_hdr.xstate_bv as '0', with out modifying the corresponding memory
|
|
|
|
* layout in the case of xsaveopt. While presenting the xstate information to
|
|
|
|
* the user, we always ensure that the memory layout of a feature will be in
|
|
|
|
* the init state if the corresponding header bit is zero. This is to ensure
|
|
|
|
* that the user doesn't see some stale state in the memory layout during
|
|
|
|
* signal handling, debugging etc.
|
|
|
|
*/
|
|
|
|
void __sanitize_i387_state(struct task_struct *tsk)
|
|
|
|
{
|
|
|
|
struct i387_fxsave_struct *fx = &tsk->thread.fpu.state->fxsave;
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
int feature_bit = 0x2;
|
|
|
|
u64 xstate_bv;
|
2010-07-20 06:05:49 +07:00
|
|
|
|
|
|
|
if (!fx)
|
|
|
|
return;
|
|
|
|
|
|
|
|
xstate_bv = tsk->thread.fpu.state->xsave.xsave_hdr.xstate_bv;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* None of the feature bits are in init state. So nothing else
|
2011-03-18 02:24:16 +07:00
|
|
|
* to do for us, as the memory layout is up to date.
|
2010-07-20 06:05:49 +07:00
|
|
|
*/
|
|
|
|
if ((xstate_bv & pcntxt_mask) == pcntxt_mask)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* FP is in init state
|
|
|
|
*/
|
|
|
|
if (!(xstate_bv & XSTATE_FP)) {
|
|
|
|
fx->cwd = 0x37f;
|
|
|
|
fx->swd = 0;
|
|
|
|
fx->twd = 0;
|
|
|
|
fx->fop = 0;
|
|
|
|
fx->rip = 0;
|
|
|
|
fx->rdp = 0;
|
|
|
|
memset(&fx->st_space[0], 0, 128);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* SSE is in init state
|
|
|
|
*/
|
|
|
|
if (!(xstate_bv & XSTATE_SSE))
|
|
|
|
memset(&fx->xmm_space[0], 0, 256);
|
|
|
|
|
|
|
|
xstate_bv = (pcntxt_mask & ~xstate_bv) >> 2;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update all the other memory layouts for which the corresponding
|
|
|
|
* header bit is in the init state.
|
|
|
|
*/
|
|
|
|
while (xstate_bv) {
|
|
|
|
if (xstate_bv & 0x1) {
|
|
|
|
int offset = xstate_offsets[feature_bit];
|
|
|
|
int size = xstate_sizes[feature_bit];
|
|
|
|
|
|
|
|
memcpy(((void *) fx) + offset,
|
|
|
|
((void *) init_xstate_buf) + offset,
|
|
|
|
size);
|
|
|
|
}
|
|
|
|
|
|
|
|
xstate_bv >>= 1;
|
|
|
|
feature_bit++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-07-30 00:29:25 +07:00
|
|
|
/*
|
|
|
|
* Check for the presence of extended state information in the
|
|
|
|
* user fpstate pointer in the sigcontext.
|
|
|
|
*/
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
static inline int check_for_xstate(struct i387_fxsave_struct __user *buf,
|
|
|
|
void __user *fpstate,
|
|
|
|
struct _fpx_sw_bytes *fx_sw)
|
2008-07-30 00:29:25 +07:00
|
|
|
{
|
|
|
|
int min_xstate_size = sizeof(struct i387_fxsave_struct) +
|
|
|
|
sizeof(struct xsave_hdr_struct);
|
|
|
|
unsigned int magic2;
|
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (__copy_from_user(fx_sw, &buf->sw_reserved[0], sizeof(*fx_sw)))
|
|
|
|
return -1;
|
2008-07-30 00:29:25 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
/* Check for the first magic field and other error scenarios. */
|
|
|
|
if (fx_sw->magic1 != FP_XSTATE_MAGIC1 ||
|
|
|
|
fx_sw->xstate_size < min_xstate_size ||
|
|
|
|
fx_sw->xstate_size > xstate_size ||
|
|
|
|
fx_sw->xstate_size > fx_sw->extended_size)
|
|
|
|
return -1;
|
2008-07-30 00:29:25 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check for the presence of second magic word at the end of memory
|
|
|
|
* layout. This detects the case where the user just copied the legacy
|
|
|
|
* fpstate layout with out copying the extended state information
|
|
|
|
* in the memory layout.
|
|
|
|
*/
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (__get_user(magic2, (__u32 __user *)(fpstate + fx_sw->xstate_size))
|
|
|
|
|| magic2 != FP_XSTATE_MAGIC2)
|
|
|
|
return -1;
|
2008-07-30 00:29:25 +07:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-07-30 00:29:22 +07:00
|
|
|
/*
|
|
|
|
* Signal frame handlers.
|
|
|
|
*/
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
static inline int save_fsave_header(struct task_struct *tsk, void __user *buf)
|
|
|
|
{
|
|
|
|
if (use_fxsr()) {
|
|
|
|
struct xsave_struct *xsave = &tsk->thread.fpu.state->xsave;
|
|
|
|
struct user_i387_ia32_struct env;
|
|
|
|
struct _fpstate_ia32 __user *fp = buf;
|
2008-07-30 00:29:22 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
convert_from_fxsr(&env, tsk);
|
|
|
|
|
|
|
|
if (__copy_to_user(buf, &env, sizeof(env)) ||
|
|
|
|
__put_user(xsave->i387.swd, &fp->status) ||
|
|
|
|
__put_user(X86_FXSR_MAGIC, &fp->magic))
|
|
|
|
return -1;
|
|
|
|
} else {
|
|
|
|
struct i387_fsave_struct __user *fp = buf;
|
|
|
|
u32 swd;
|
|
|
|
if (__get_user(swd, &fp->swd) || __put_user(swd, &fp->status))
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int save_xstate_epilog(void __user *buf, int ia32_frame)
|
2008-07-30 00:29:22 +07:00
|
|
|
{
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
struct xsave_struct __user *x = buf;
|
|
|
|
struct _fpx_sw_bytes *sw_bytes;
|
|
|
|
u32 xstate_bv;
|
|
|
|
int err;
|
2008-07-30 00:29:22 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
/* Setup the bytes not touched by the [f]xsave and reserved for SW. */
|
|
|
|
sw_bytes = ia32_frame ? &fx_sw_reserved_ia32 : &fx_sw_reserved;
|
|
|
|
err = __copy_to_user(&x->i387.sw_reserved, sw_bytes, sizeof(*sw_bytes));
|
2008-07-30 00:29:22 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (!use_xsave())
|
|
|
|
return err;
|
2008-07-30 00:29:22 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
err |= __put_user(FP_XSTATE_MAGIC2, (__u32 *)(buf + xstate_size));
|
2008-07-30 00:29:22 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
/*
|
|
|
|
* Read the xstate_bv which we copied (directly from the cpu or
|
|
|
|
* from the state in task struct) to the user buffers.
|
|
|
|
*/
|
|
|
|
err |= __get_user(xstate_bv, (__u32 *)&x->xsave_hdr.xstate_bv);
|
2009-04-10 05:24:34 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
/*
|
|
|
|
* For legacy compatible, we always set FP/SSE bits in the bit
|
|
|
|
* vector while saving the state to the user context. This will
|
|
|
|
* enable us capturing any changes(during sigreturn) to
|
|
|
|
* the FP/SSE bits by the legacy applications which don't touch
|
|
|
|
* xstate_bv in the xsave header.
|
|
|
|
*
|
|
|
|
* xsave aware apps can change the xstate_bv in the xsave
|
|
|
|
* header as well as change any contents in the memory layout.
|
|
|
|
* xrestore as part of sigreturn will capture all the changes.
|
|
|
|
*/
|
|
|
|
xstate_bv |= XSTATE_FPSSE;
|
2008-07-30 00:29:25 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
err |= __put_user(xstate_bv, (__u32 *)&x->xsave_hdr.xstate_bv);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int save_user_xstate(struct xsave_struct __user *buf)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (use_xsave())
|
|
|
|
err = xsave_user(buf);
|
|
|
|
else if (use_fxsr())
|
|
|
|
err = fxsave_user((struct i387_fxsave_struct __user *) buf);
|
|
|
|
else
|
|
|
|
err = fsave_user((struct i387_fsave_struct __user *) buf);
|
|
|
|
|
|
|
|
if (unlikely(err) && __clear_user(buf, xstate_size))
|
|
|
|
err = -EFAULT;
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the fpu, extended register state to the user signal frame.
|
|
|
|
*
|
|
|
|
* 'buf_fx' is the 64-byte aligned pointer at which the [f|fx|x]save
|
|
|
|
* state is copied.
|
|
|
|
* 'buf' points to the 'buf_fx' or to the fsave header followed by 'buf_fx'.
|
|
|
|
*
|
|
|
|
* buf == buf_fx for 64-bit frames and 32-bit fsave frame.
|
|
|
|
* buf != buf_fx for 32-bit frames with fxstate.
|
|
|
|
*
|
|
|
|
* If the fpu, extended register state is live, save the state directly
|
|
|
|
* to the user frame pointed by the aligned pointer 'buf_fx'. Otherwise,
|
|
|
|
* copy the thread's fpu state to the user frame starting at 'buf_fx'.
|
|
|
|
*
|
|
|
|
* If this is a 32-bit frame with fxstate, put a fsave header before
|
|
|
|
* the aligned state at 'buf_fx'.
|
|
|
|
*
|
|
|
|
* For [f]xsave state, update the SW reserved fields in the [f]xsave frame
|
|
|
|
* indicating the absence/presence of the extended state to the user.
|
|
|
|
*/
|
|
|
|
int save_xstate_sig(void __user *buf, void __user *buf_fx, int size)
|
|
|
|
{
|
|
|
|
struct xsave_struct *xsave = ¤t->thread.fpu.state->xsave;
|
|
|
|
struct task_struct *tsk = current;
|
|
|
|
int ia32_fxstate = (buf != buf_fx);
|
|
|
|
|
|
|
|
ia32_fxstate &= (config_enabled(CONFIG_X86_32) ||
|
|
|
|
config_enabled(CONFIG_IA32_EMULATION));
|
|
|
|
|
|
|
|
if (!access_ok(VERIFY_WRITE, buf, size))
|
|
|
|
return -EACCES;
|
|
|
|
|
2013-04-29 21:04:20 +07:00
|
|
|
if (!static_cpu_has(X86_FEATURE_FPU))
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
return fpregs_soft_get(current, NULL, 0,
|
|
|
|
sizeof(struct user_i387_ia32_struct), NULL,
|
|
|
|
(struct _fpstate_ia32 __user *) buf) ? -1 : 1;
|
|
|
|
|
|
|
|
if (user_has_fpu()) {
|
|
|
|
/* Save the live register state to the user directly. */
|
|
|
|
if (save_user_xstate(buf_fx))
|
|
|
|
return -1;
|
|
|
|
/* Update the thread's fxstate to save the fsave header. */
|
|
|
|
if (ia32_fxstate)
|
|
|
|
fpu_fxsave(&tsk->thread.fpu);
|
2008-07-30 00:29:22 +07:00
|
|
|
} else {
|
2010-07-20 06:05:49 +07:00
|
|
|
sanitize_i387_state(tsk);
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (__copy_to_user(buf_fx, xsave, xstate_size))
|
2008-07-30 00:29:22 +07:00
|
|
|
return -1;
|
|
|
|
}
|
2008-07-30 00:29:25 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
/* Save the fsave header for the 32-bit frames. */
|
|
|
|
if ((ia32_fxstate || !use_fxsr()) && save_fsave_header(tsk, buf))
|
|
|
|
return -1;
|
2009-04-10 05:24:34 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (use_fxsr() && save_xstate_epilog(buf_fx, ia32_fxstate))
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2008-07-30 00:29:25 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
static inline void
|
|
|
|
sanitize_restored_xstate(struct task_struct *tsk,
|
|
|
|
struct user_i387_ia32_struct *ia32_env,
|
|
|
|
u64 xstate_bv, int fx_only)
|
|
|
|
{
|
|
|
|
struct xsave_struct *xsave = &tsk->thread.fpu.state->xsave;
|
|
|
|
struct xsave_hdr_struct *xsave_hdr = &xsave->xsave_hdr;
|
2008-07-30 00:29:25 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (use_xsave()) {
|
|
|
|
/* These bits must be zero. */
|
2014-05-30 01:12:43 +07:00
|
|
|
memset(xsave_hdr->reserved, 0, 48);
|
2008-10-08 04:04:28 +07:00
|
|
|
|
|
|
|
/*
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
* Init the state that is not present in the memory
|
|
|
|
* layout and not enabled by the OS.
|
2008-10-08 04:04:28 +07:00
|
|
|
*/
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (fx_only)
|
|
|
|
xsave_hdr->xstate_bv = XSTATE_FPSSE;
|
|
|
|
else
|
|
|
|
xsave_hdr->xstate_bv &= (pcntxt_mask & xstate_bv);
|
|
|
|
}
|
2008-10-08 04:04:28 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (use_fxsr()) {
|
2008-10-08 04:04:28 +07:00
|
|
|
/*
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
* mscsr reserved bits must be masked to zero for security
|
|
|
|
* reasons.
|
2008-10-08 04:04:28 +07:00
|
|
|
*/
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
xsave->i387.mxcsr &= mxcsr_feature_mask;
|
2008-10-08 04:04:28 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
convert_to_fxsr(tsk, ia32_env);
|
2008-07-30 00:29:25 +07:00
|
|
|
}
|
2008-07-30 00:29:22 +07:00
|
|
|
}
|
|
|
|
|
2008-07-30 00:29:25 +07:00
|
|
|
/*
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
* Restore the extended state if present. Otherwise, restore the FP/SSE state.
|
2008-07-30 00:29:25 +07:00
|
|
|
*/
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
static inline int restore_user_xstate(void __user *buf, u64 xbv, int fx_only)
|
2008-07-30 00:29:25 +07:00
|
|
|
{
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (use_xsave()) {
|
|
|
|
if ((unsigned long)buf % 64 || fx_only) {
|
|
|
|
u64 init_bv = pcntxt_mask & ~XSTATE_FPSSE;
|
|
|
|
xrstor_state(init_xstate_buf, init_bv);
|
2012-09-26 05:42:18 +07:00
|
|
|
return fxrstor_user(buf);
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
} else {
|
|
|
|
u64 init_bv = pcntxt_mask & ~xbv;
|
|
|
|
if (unlikely(init_bv))
|
|
|
|
xrstor_state(init_xstate_buf, init_bv);
|
|
|
|
return xrestore_user(buf, xbv);
|
|
|
|
}
|
|
|
|
} else if (use_fxsr()) {
|
2012-09-26 05:42:18 +07:00
|
|
|
return fxrstor_user(buf);
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
} else
|
2012-09-26 05:42:18 +07:00
|
|
|
return frstor_user(buf);
|
2008-07-30 00:29:25 +07:00
|
|
|
}
|
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
int __restore_xstate_sig(void __user *buf, void __user *buf_fx, int size)
|
2008-07-30 00:29:22 +07:00
|
|
|
{
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
int ia32_fxstate = (buf != buf_fx);
|
2008-07-30 00:29:22 +07:00
|
|
|
struct task_struct *tsk = current;
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
int state_size = xstate_size;
|
|
|
|
u64 xstate_bv = 0;
|
|
|
|
int fx_only = 0;
|
|
|
|
|
|
|
|
ia32_fxstate &= (config_enabled(CONFIG_X86_32) ||
|
|
|
|
config_enabled(CONFIG_IA32_EMULATION));
|
2008-07-30 00:29:22 +07:00
|
|
|
|
|
|
|
if (!buf) {
|
2015-03-16 16:21:55 +07:00
|
|
|
fpu_reset_state(tsk);
|
2008-07-30 00:29:22 +07:00
|
|
|
return 0;
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!access_ok(VERIFY_READ, buf, size))
|
|
|
|
return -EACCES;
|
|
|
|
|
2015-04-03 17:02:02 +07:00
|
|
|
if (!used_math() && fpstate_alloc_init(tsk))
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
return -1;
|
2008-07-30 00:29:22 +07:00
|
|
|
|
2013-04-29 21:04:20 +07:00
|
|
|
if (!static_cpu_has(X86_FEATURE_FPU))
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
return fpregs_soft_set(current, NULL,
|
|
|
|
0, sizeof(struct user_i387_ia32_struct),
|
|
|
|
NULL, buf) != 0;
|
2008-07-30 00:29:22 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (use_xsave()) {
|
|
|
|
struct _fpx_sw_bytes fx_sw_user;
|
|
|
|
if (unlikely(check_for_xstate(buf_fx, buf_fx, &fx_sw_user))) {
|
|
|
|
/*
|
|
|
|
* Couldn't find the extended state information in the
|
|
|
|
* memory layout. Restore just the FP/SSE and init all
|
|
|
|
* the other extended state.
|
|
|
|
*/
|
|
|
|
state_size = sizeof(struct i387_fxsave_struct);
|
|
|
|
fx_only = 1;
|
|
|
|
} else {
|
|
|
|
state_size = fx_sw_user.xstate_size;
|
|
|
|
xstate_bv = fx_sw_user.xstate_bv;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (ia32_fxstate) {
|
|
|
|
/*
|
|
|
|
* For 32-bit frames with fxstate, copy the user state to the
|
|
|
|
* thread's fpu state, reconstruct fxstate from the fsave
|
|
|
|
* header. Sanitize the copied state etc.
|
|
|
|
*/
|
2015-03-13 15:53:09 +07:00
|
|
|
struct fpu *fpu = &tsk->thread.fpu;
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
struct user_i387_ia32_struct env;
|
2012-08-25 04:13:02 +07:00
|
|
|
int err = 0;
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
|
2012-08-25 04:13:02 +07:00
|
|
|
/*
|
|
|
|
* Drop the current fpu which clears used_math(). This ensures
|
|
|
|
* that any context-switch during the copy of the new state,
|
|
|
|
* avoids the intermediate state from getting restored/saved.
|
|
|
|
* Thus avoiding the new restored state from getting corrupted.
|
|
|
|
* We will be ready to restore/save the state only after
|
|
|
|
* set_used_math() is again set.
|
|
|
|
*/
|
2012-08-25 04:12:57 +07:00
|
|
|
drop_fpu(tsk);
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
|
2015-03-13 15:53:09 +07:00
|
|
|
if (__copy_from_user(&fpu->state->xsave, buf_fx, state_size) ||
|
2012-08-25 04:13:02 +07:00
|
|
|
__copy_from_user(&env, buf, sizeof(env))) {
|
2015-04-03 18:01:52 +07:00
|
|
|
fpstate_init(fpu);
|
2012-08-25 04:13:02 +07:00
|
|
|
err = -1;
|
|
|
|
} else {
|
|
|
|
sanitize_restored_xstate(tsk, &env, xstate_bv, fx_only);
|
|
|
|
}
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
|
2015-03-13 15:53:09 +07:00
|
|
|
set_used_math();
|
2014-09-03 00:57:17 +07:00
|
|
|
if (use_eager_fpu()) {
|
|
|
|
preempt_disable();
|
2015-04-22 18:16:47 +07:00
|
|
|
fpu__restore();
|
2014-09-03 00:57:17 +07:00
|
|
|
preempt_enable();
|
|
|
|
}
|
2012-08-25 04:13:02 +07:00
|
|
|
|
|
|
|
return err;
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
} else {
|
2008-07-30 00:29:22 +07:00
|
|
|
/*
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
* For 64-bit frames and 32-bit fsave frames, restore the user
|
|
|
|
* state to the registers directly (with exceptions handled).
|
2008-07-30 00:29:22 +07:00
|
|
|
*/
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
user_fpu_begin();
|
|
|
|
if (restore_user_xstate(buf_fx, xstate_bv, fx_only)) {
|
2015-03-16 16:21:55 +07:00
|
|
|
fpu_reset_state(tsk);
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
return -1;
|
|
|
|
}
|
2008-07-30 00:29:22 +07:00
|
|
|
}
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
|
|
|
|
return 0;
|
2008-07-30 00:29:22 +07:00
|
|
|
}
|
|
|
|
|
2008-07-30 00:29:25 +07:00
|
|
|
/*
|
|
|
|
* Prepare the SW reserved portion of the fxsave memory layout, indicating
|
|
|
|
* the presence of the extended state information in the memory layout
|
|
|
|
* pointed by the fpstate pointer in the sigcontext.
|
|
|
|
* This will be saved when ever the FP and extended state context is
|
|
|
|
* saved on the user stack during the signal handler delivery to the user.
|
|
|
|
*/
|
2008-10-22 06:49:09 +07:00
|
|
|
static void prepare_fx_sw_frame(void)
|
2008-07-30 00:29:25 +07:00
|
|
|
{
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
int fsave_header_size = sizeof(struct i387_fsave_struct);
|
|
|
|
int size = xstate_size + FP_XSTATE_MAGIC2_SIZE;
|
2008-07-30 00:29:25 +07:00
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (config_enabled(CONFIG_X86_32))
|
|
|
|
size += fsave_header_size;
|
2008-07-30 00:29:25 +07:00
|
|
|
|
|
|
|
fx_sw_reserved.magic1 = FP_XSTATE_MAGIC1;
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
fx_sw_reserved.extended_size = size;
|
2008-07-30 07:23:16 +07:00
|
|
|
fx_sw_reserved.xstate_bv = pcntxt_mask;
|
2008-07-30 00:29:25 +07:00
|
|
|
fx_sw_reserved.xstate_size = xstate_size;
|
|
|
|
|
x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2012-07-25 06:05:29 +07:00
|
|
|
if (config_enabled(CONFIG_IA32_EMULATION)) {
|
|
|
|
fx_sw_reserved_ia32 = fx_sw_reserved;
|
|
|
|
fx_sw_reserved_ia32.extended_size += fsave_header_size;
|
|
|
|
}
|
|
|
|
}
|
2008-07-30 00:29:21 +07:00
|
|
|
|
2008-07-30 00:29:19 +07:00
|
|
|
/*
|
|
|
|
* Enable the extended processor state save/restore feature
|
|
|
|
*/
|
2010-07-22 04:23:10 +07:00
|
|
|
static inline void xstate_enable(void)
|
2008-07-30 00:29:19 +07:00
|
|
|
{
|
2014-10-25 05:58:07 +07:00
|
|
|
cr4_set_bits(X86_CR4_OSXSAVE);
|
2008-07-30 07:23:16 +07:00
|
|
|
xsetbv(XCR_XFEATURE_ENABLED_MASK, pcntxt_mask);
|
2008-07-30 00:29:19 +07:00
|
|
|
}
|
|
|
|
|
2010-07-20 06:05:48 +07:00
|
|
|
/*
|
|
|
|
* Record the offsets and sizes of different state managed by the xsave
|
|
|
|
* memory layout.
|
|
|
|
*/
|
2010-07-22 00:03:56 +07:00
|
|
|
static void __init setup_xstate_features(void)
|
2010-07-20 06:05:48 +07:00
|
|
|
{
|
|
|
|
int eax, ebx, ecx, edx, leaf = 0x2;
|
|
|
|
|
|
|
|
xstate_features = fls64(pcntxt_mask);
|
|
|
|
xstate_offsets = alloc_bootmem(xstate_features * sizeof(int));
|
|
|
|
xstate_sizes = alloc_bootmem(xstate_features * sizeof(int));
|
|
|
|
|
|
|
|
do {
|
2010-07-22 00:03:54 +07:00
|
|
|
cpuid_count(XSTATE_CPUID, leaf, &eax, &ebx, &ecx, &edx);
|
2010-07-20 06:05:48 +07:00
|
|
|
|
|
|
|
if (eax == 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
xstate_offsets[leaf] = ebx;
|
|
|
|
xstate_sizes[leaf] = eax;
|
|
|
|
|
|
|
|
leaf++;
|
|
|
|
} while (1);
|
|
|
|
}
|
|
|
|
|
2014-05-30 01:12:44 +07:00
|
|
|
/*
|
|
|
|
* This function sets up offsets and sizes of all extended states in
|
|
|
|
* xsave area. This supports both standard format and compacted format
|
|
|
|
* of the xsave aread.
|
|
|
|
*
|
|
|
|
* Input: void
|
|
|
|
* Output: void
|
|
|
|
*/
|
|
|
|
void setup_xstate_comp(void)
|
|
|
|
{
|
2014-05-31 04:59:24 +07:00
|
|
|
unsigned int xstate_comp_sizes[sizeof(pcntxt_mask)*8];
|
2014-05-30 01:12:44 +07:00
|
|
|
int i;
|
|
|
|
|
2014-05-31 04:59:24 +07:00
|
|
|
/*
|
|
|
|
* The FP xstates and SSE xstates are legacy states. They are always
|
|
|
|
* in the fixed offsets in the xsave area in either compacted form
|
|
|
|
* or standard form.
|
|
|
|
*/
|
|
|
|
xstate_comp_offsets[0] = 0;
|
|
|
|
xstate_comp_offsets[1] = offsetof(struct i387_fxsave_struct, xmm_space);
|
2014-05-30 01:12:44 +07:00
|
|
|
|
|
|
|
if (!cpu_has_xsaves) {
|
|
|
|
for (i = 2; i < xstate_features; i++) {
|
|
|
|
if (test_bit(i, (unsigned long *)&pcntxt_mask)) {
|
|
|
|
xstate_comp_offsets[i] = xstate_offsets[i];
|
|
|
|
xstate_comp_sizes[i] = xstate_sizes[i];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
xstate_comp_offsets[2] = FXSAVE_SIZE + XSAVE_HDR_SIZE;
|
|
|
|
|
|
|
|
for (i = 2; i < xstate_features; i++) {
|
|
|
|
if (test_bit(i, (unsigned long *)&pcntxt_mask))
|
|
|
|
xstate_comp_sizes[i] = xstate_sizes[i];
|
|
|
|
else
|
|
|
|
xstate_comp_sizes[i] = 0;
|
|
|
|
|
|
|
|
if (i > 2)
|
|
|
|
xstate_comp_offsets[i] = xstate_comp_offsets[i-1]
|
|
|
|
+ xstate_comp_sizes[i-1];
|
|
|
|
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-07-30 00:29:19 +07:00
|
|
|
/*
|
|
|
|
* setup the xstate image representing the init state
|
|
|
|
*/
|
2012-09-07 04:58:52 +07:00
|
|
|
static void __init setup_init_fpu_buf(void)
|
2008-07-30 00:29:19 +07:00
|
|
|
{
|
2010-07-20 06:05:49 +07:00
|
|
|
/*
|
|
|
|
* Setup init_xstate_buf to represent the init state of
|
|
|
|
* all the features managed by the xsave
|
|
|
|
*/
|
2010-11-17 04:23:51 +07:00
|
|
|
init_xstate_buf = alloc_bootmem_align(xstate_size,
|
|
|
|
__alignof__(struct xsave_struct));
|
2012-09-07 04:58:52 +07:00
|
|
|
fx_finit(&init_xstate_buf->i387);
|
|
|
|
|
|
|
|
if (!cpu_has_xsave)
|
|
|
|
return;
|
|
|
|
|
|
|
|
setup_xstate_features();
|
2010-07-20 06:05:48 +07:00
|
|
|
|
2014-05-30 01:12:42 +07:00
|
|
|
if (cpu_has_xsaves) {
|
|
|
|
init_xstate_buf->xsave_hdr.xcomp_bv =
|
|
|
|
(u64)1 << 63 | pcntxt_mask;
|
|
|
|
init_xstate_buf->xsave_hdr.xstate_bv = pcntxt_mask;
|
|
|
|
}
|
|
|
|
|
2010-07-20 06:05:49 +07:00
|
|
|
/*
|
|
|
|
* Init all the features state with header_bv being 0x0
|
|
|
|
*/
|
2014-05-30 01:12:42 +07:00
|
|
|
xrstor_state_booting(init_xstate_buf, -1);
|
2010-07-20 06:05:49 +07:00
|
|
|
/*
|
|
|
|
* Dump the init state again. This is to identify the init state
|
|
|
|
* of any feature which is not represented by all zero's.
|
|
|
|
*/
|
2014-05-30 01:12:42 +07:00
|
|
|
xsave_state_booting(init_xstate_buf, -1);
|
2008-07-30 00:29:19 +07:00
|
|
|
}
|
|
|
|
|
2012-09-11 00:32:32 +07:00
|
|
|
static enum { AUTO, ENABLE, DISABLE } eagerfpu = AUTO;
|
2012-09-07 04:58:52 +07:00
|
|
|
static int __init eager_fpu_setup(char *s)
|
|
|
|
{
|
|
|
|
if (!strcmp(s, "on"))
|
2012-09-11 00:32:32 +07:00
|
|
|
eagerfpu = ENABLE;
|
2012-09-07 04:58:52 +07:00
|
|
|
else if (!strcmp(s, "off"))
|
2012-09-11 00:32:32 +07:00
|
|
|
eagerfpu = DISABLE;
|
|
|
|
else if (!strcmp(s, "auto"))
|
|
|
|
eagerfpu = AUTO;
|
2012-09-07 04:58:52 +07:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("eagerfpu=", eager_fpu_setup);
|
|
|
|
|
2014-05-30 01:12:43 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Calculate total size of enabled xstates in XCR0/pcntxt_mask.
|
|
|
|
*/
|
|
|
|
static void __init init_xstate_size(void)
|
|
|
|
{
|
|
|
|
unsigned int eax, ebx, ecx, edx;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (!cpu_has_xsaves) {
|
|
|
|
cpuid_count(XSTATE_CPUID, 0, &eax, &ebx, &ecx, &edx);
|
|
|
|
xstate_size = ebx;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
xstate_size = FXSAVE_SIZE + XSAVE_HDR_SIZE;
|
|
|
|
for (i = 2; i < 64; i++) {
|
|
|
|
if (test_bit(i, (unsigned long *)&pcntxt_mask)) {
|
|
|
|
cpuid_count(XSTATE_CPUID, i, &eax, &ebx, &ecx, &edx);
|
|
|
|
xstate_size += eax;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-07-30 00:29:19 +07:00
|
|
|
/*
|
|
|
|
* Enable and initialize the xsave feature.
|
|
|
|
*/
|
2010-07-22 04:23:10 +07:00
|
|
|
static void __init xstate_enable_boot_cpu(void)
|
2008-07-30 00:29:19 +07:00
|
|
|
{
|
|
|
|
unsigned int eax, ebx, ecx, edx;
|
|
|
|
|
2010-07-22 00:03:54 +07:00
|
|
|
if (boot_cpu_data.cpuid_level < XSTATE_CPUID) {
|
|
|
|
WARN(1, KERN_ERR "XSTATE_CPUID missing\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
cpuid_count(XSTATE_CPUID, 0, &eax, &ebx, &ecx, &edx);
|
2008-07-30 07:23:16 +07:00
|
|
|
pcntxt_mask = eax + ((u64)edx << 32);
|
2008-07-30 00:29:19 +07:00
|
|
|
|
2008-07-30 07:23:16 +07:00
|
|
|
if ((pcntxt_mask & XSTATE_FPSSE) != XSTATE_FPSSE) {
|
2012-05-22 09:50:07 +07:00
|
|
|
pr_err("FP/SSE not shown under xsave features 0x%llx\n",
|
2008-07-30 07:23:16 +07:00
|
|
|
pcntxt_mask);
|
2008-07-30 00:29:19 +07:00
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-04-11 05:21:24 +07:00
|
|
|
* Support only the state known to OS.
|
2008-07-30 00:29:19 +07:00
|
|
|
*/
|
2008-07-30 07:23:16 +07:00
|
|
|
pcntxt_mask = pcntxt_mask & XCNTXT_MASK;
|
2010-07-22 00:03:53 +07:00
|
|
|
|
2010-07-22 04:23:10 +07:00
|
|
|
xstate_enable();
|
2008-07-30 00:29:19 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Recompute the context size for enabled features
|
|
|
|
*/
|
2014-05-30 01:12:43 +07:00
|
|
|
init_xstate_size();
|
2008-07-30 00:29:19 +07:00
|
|
|
|
2010-02-12 02:50:59 +07:00
|
|
|
update_regset_xstate_info(xstate_size, pcntxt_mask);
|
2008-07-30 00:29:25 +07:00
|
|
|
prepare_fx_sw_frame();
|
2012-09-07 04:58:52 +07:00
|
|
|
setup_init_fpu_buf();
|
2008-07-30 00:29:19 +07:00
|
|
|
|
2012-09-11 00:32:32 +07:00
|
|
|
/* Auto enable eagerfpu for xsaveopt */
|
|
|
|
if (cpu_has_xsaveopt && eagerfpu != DISABLE)
|
|
|
|
eagerfpu = ENABLE;
|
2012-09-07 05:05:18 +07:00
|
|
|
|
2013-12-05 16:15:34 +07:00
|
|
|
if (pcntxt_mask & XSTATE_EAGER) {
|
|
|
|
if (eagerfpu == DISABLE) {
|
|
|
|
pr_err("eagerfpu not present, disabling some xstate features: 0x%llx\n",
|
|
|
|
pcntxt_mask & XSTATE_EAGER);
|
|
|
|
pcntxt_mask &= ~XSTATE_EAGER;
|
|
|
|
} else {
|
|
|
|
eagerfpu = ENABLE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-05-30 01:12:43 +07:00
|
|
|
pr_info("enabled xstate_bv 0x%llx, cntxt size 0x%x using %s\n",
|
|
|
|
pcntxt_mask, xstate_size,
|
|
|
|
cpu_has_xsaves ? "compacted form" : "standard form");
|
2008-07-30 00:29:19 +07:00
|
|
|
}
|
2010-07-21 01:50:51 +07:00
|
|
|
|
2010-07-22 04:23:10 +07:00
|
|
|
/*
|
|
|
|
* For the very first instance, this calls xstate_enable_boot_cpu();
|
|
|
|
* for all subsequent instances, this calls xstate_enable().
|
|
|
|
*
|
|
|
|
* This is somewhat obfuscated due to the lack of powerful enough
|
|
|
|
* overrides for the section checks.
|
|
|
|
*/
|
x86: delete __cpuinit usage from all x86 files
The __cpuinit type of throwaway sections might have made sense
some time ago when RAM was more constrained, but now the savings
do not offset the cost and complications. For example, the fix in
commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time")
is a good example of the nasty type of bugs that can be created
with improper use of the various __init prefixes.
After a discussion on LKML[1] it was decided that cpuinit should go
the way of devinit and be phased out. Once all the users are gone,
we can then finally remove the macros themselves from linux/init.h.
Note that some harmless section mismatch warnings may result, since
notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c)
are flagged as __cpuinit -- so if we remove the __cpuinit from
arch specific callers, we will also get section mismatch warnings.
As an intermediate step, we intend to turn the linux/init.h cpuinit
content into no-ops as early as possible, since that will get rid
of these warnings. In any case, they are temporary and harmless.
This removes all the arch/x86 uses of the __cpuinit macros from
all C files. x86 only had the one __CPUINIT used in assembly files,
and it wasn't paired off with a .previous or a __FINIT, so we can
delete it directly w/o any corresponding additional change there.
[1] https://lkml.org/lkml/2013/5/20/589
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: x86@kernel.org
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: H. Peter Anvin <hpa@linux.intel.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-19 05:23:59 +07:00
|
|
|
void xsave_init(void)
|
2010-07-21 01:50:51 +07:00
|
|
|
{
|
2010-07-22 04:23:10 +07:00
|
|
|
static __refdata void (*next_func)(void) = xstate_enable_boot_cpu;
|
|
|
|
void (*this_func)(void);
|
|
|
|
|
2010-07-22 00:03:52 +07:00
|
|
|
if (!cpu_has_xsave)
|
|
|
|
return;
|
|
|
|
|
2010-07-22 04:23:10 +07:00
|
|
|
this_func = next_func;
|
2012-09-07 04:58:52 +07:00
|
|
|
next_func = xstate_enable;
|
2010-07-22 04:23:10 +07:00
|
|
|
this_func();
|
2010-07-21 01:50:51 +07:00
|
|
|
}
|
2012-09-07 04:58:52 +07:00
|
|
|
|
2015-03-14 22:13:34 +07:00
|
|
|
/*
|
|
|
|
* setup_init_fpu_buf() is __init and it is OK to call it here because
|
|
|
|
* init_xstate_buf will be unset only once during boot.
|
|
|
|
*/
|
|
|
|
void __init_refok eager_fpu_init(void)
|
2012-09-07 04:58:52 +07:00
|
|
|
{
|
2015-01-20 01:52:12 +07:00
|
|
|
WARN_ON(used_math());
|
2012-09-07 04:58:52 +07:00
|
|
|
current_thread_info()->status = 0;
|
2012-09-11 00:32:32 +07:00
|
|
|
|
|
|
|
if (eagerfpu == ENABLE)
|
|
|
|
setup_force_cpu_cap(X86_FEATURE_EAGER_FPU);
|
|
|
|
|
2012-09-07 04:58:52 +07:00
|
|
|
if (!cpu_has_eager_fpu) {
|
|
|
|
stts();
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2015-03-14 22:13:34 +07:00
|
|
|
if (!init_xstate_buf)
|
|
|
|
setup_init_fpu_buf();
|
2012-09-07 04:58:52 +07:00
|
|
|
}
|
2014-05-30 01:12:44 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Given the xsave area and a state inside, this function returns the
|
|
|
|
* address of the state.
|
|
|
|
*
|
|
|
|
* This is the API that is called to get xstate address in either
|
|
|
|
* standard format or compacted format of xsave area.
|
|
|
|
*
|
|
|
|
* Inputs:
|
|
|
|
* xsave: base address of the xsave area;
|
|
|
|
* xstate: state which is defined in xsave.h (e.g. XSTATE_FP, XSTATE_SSE,
|
|
|
|
* etc.)
|
|
|
|
* Output:
|
|
|
|
* address of the state in the xsave area.
|
|
|
|
*/
|
|
|
|
void *get_xsave_addr(struct xsave_struct *xsave, int xstate)
|
|
|
|
{
|
|
|
|
int feature = fls64(xstate) - 1;
|
|
|
|
if (!test_bit(feature, (unsigned long *)&pcntxt_mask))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return (void *)xsave + xstate_comp_offsets[feature];
|
|
|
|
}
|
2014-11-24 16:57:42 +07:00
|
|
|
EXPORT_SYMBOL_GPL(get_xsave_addr);
|