module_trampoline_target() is only used by ftrace. Move the prototype
within the appropriate #ifdef in the header. Also, move the function
body to the end of module_64.c so as to consolidate all ftrace code in
one place.
No functional changes.
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/2527351f65c53c5866068ae130dc34c5d4ee8ad9.1587488954.git.naveen.n.rao@linux.vnet.ibm.com
Mapping of early shadow area is implemented by using a single static
page table having all entries pointing to the same early shadow page.
The shadow area must therefore occupy full PGD entries.
The shadow area has a size of 128MB starting at 0xf8000000.
With 4k pages, a PGD entry is 4MB
With 16k pages, a PGD entry is 64MB
With 64k pages, a PGD entry is 1GB which is too big.
Until we rework the early shadow mapping, disable KASAN when the page
size is too big.
Fixes: 2edb16efc8 ("powerpc/32: Add KASAN support")
Cc: stable@vger.kernel.org # v5.2+
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/7195fcde7314ccbf7a081b356084a69d421b10d4.1590660977.git.christophe.leroy@csgroup.eu
On book3s/32, KUEP is an heavy process as it requires to
set/unset the NX bit in each of the 12 user segments
everytime the kernel is entered/exited from/to user space.
Don't select KUEP by default on book3s/32.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1492bb150c1aaa53d99a604b49992e60ea20cd5f.1586962582.git.christophe.leroy@c-s.fr
On book3s/32, KUAP is an heavy process as it requires to
determine which segments are impacted and unlock/lock
each of them.
And since the implementation of user_access_begin/end, it
is even worth for the time being because unlike __get_user(),
user_access_begin doesn't make difference between read and write
and unlocks access also for read allthought that's unneeded
on book3s/32.
As shown by the size of a kernel built with KUAP and one without,
the overhead is 64k bytes of code. As a comparison a similar
build on an 8xx has an overhead of only 8k bytes of code.
text data bss dec hex filename
7230416 1425868 837376 9493660 90dc9c vmlinux.kuap6xx
7165012 1425548 837376 9427936 8fdbe0 vmlinux.nokuap6xx
6519796 1960028 477464 8957288 88ad68 vmlinux.kuap8xx
6511664 1959864 477464 8948992 888d00 vmlinux.nokuap8xx
Until a more optimised KUAP is implemented on book3s/32,
don't select it by default.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/154a99399317b096ac1f04827b9f8d7a9179ddc1.1586962586.git.christophe.leroy@c-s.fr
To enable/disable kernel access to user space, the 8xx has to
modify the properties of access group 1. This is done by writing
predefined values into SPRN_Mx_AP registers.
As of today, a __put_user() gives:
00000d64 <my_test>:
d64: 3d 20 4f ff lis r9,20479
d68: 61 29 ff ff ori r9,r9,65535
d6c: 7d 3a c3 a6 mtspr 794,r9
d70: 39 20 00 00 li r9,0
d74: 90 83 00 00 stw r4,0(r3)
d78: 3d 20 6f ff lis r9,28671
d7c: 61 29 ff ff ori r9,r9,65535
d80: 7d 3a c3 a6 mtspr 794,r9
d84: 4e 80 00 20 blr
Because only groups 0 and 1 are used, the definition of
groups 2 to 15 doesn't matter.
By setting unused bits to 0 instead on 1, one instruction is
removed for each lock and unlock action:
00000d5c <my_test>:
d5c: 3d 20 40 00 lis r9,16384
d60: 7d 3a c3 a6 mtspr 794,r9
d64: 39 20 00 00 li r9,0
d68: 90 83 00 00 stw r4,0(r3)
d6c: 3d 20 60 00 lis r9,24576
d70: 7d 3a c3 a6 mtspr 794,r9
d74: 4e 80 00 20 blr
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/57425c33dd72f292b1a23570244b81419072a7aa.1586945153.git.christophe.leroy@c-s.fr
kprobe does not handle events happening in real mode.
The very last part of exception exits cannot support a trap.
Blacklist them from kprobe.
While we are at it, remove exc_exit_start symbol which is not
used to avoid having to blacklist it.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/098b0fd3f6299aa1bd692bd576bd7012c84608de.1585670437.git.christophe.leroy@c-s.fr
kprobe does not handle events happening in real mode.
The very last part of syscall cannot support a trap.
Add a symbol syscall_exit_finish to identify that part and
blacklist it from kprobe.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/23eddf49abb03d1359fa0be4206998eb3800f42c.1585670437.git.christophe.leroy@c-s.fr
kprobe does not handle events happening in real mode.
As exception entry points are running with MMU disabled,
blacklist them.
The handling of TLF_NAPPING and TLF_SLEEPING is moved before the
CONFIG_TRACE_IRQFLAGS which contains 'reenable_mmu' because from there
kprobe will be possible as the kernel will run with MMU enabled.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/f61ac599855e674ebb592464d0ea32a3ba9c6644.1585670437.git.christophe.leroy@c-s.fr
In hash_low.S, a lot of named local symbols are used instead of
numbers to ease code readability. However, they don't need to be
visible.
In order to ease blacklisting of functions running with MMU
disabled for kprobe, rename the symbols to .Lsymbols in order
to hide them as if they were numbered labels.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/90c430d9e0f7af772a58aaeaf17bcc6321265340.1585670437.git.christophe.leroy@c-s.fr
This gives us OF_PMEM which is useful in mambo.
This adds 153K to the text of ppc64le_defconfig which 0.8% of the
total text.
LIBNVDIMM text data bss dec hex
Without 18574833 5518150 1539240 25632223 1871ddf
With 18727834 5546206 1539368 25813408 189e1a0
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200519043009.3081885-1-mikey@neuling.org
Implement rtas_call_reentrant() for reentrant rtas-calls:
"ibm,int-on", "ibm,int-off",ibm,get-xive" and "ibm,set-xive".
On LoPAPR Version 1.1 (March 24, 2016), from 7.3.10.1 to 7.3.10.4,
items 2 and 3 say:
2 - For the PowerPC External Interrupt option: The * call must be
reentrant to the number of processors on the platform.
3 - For the PowerPC External Interrupt option: The * argument call
buffer for each simultaneous call must be physically unique.
So, these rtas-calls can be called in a lockless way, if using
a different buffer for each cpu doing such rtas call.
For this, it was suggested to add the buffer (struct rtas_args)
in the PACA struct, so each cpu can have it's own buffer.
The PACA struct received a pointer to rtas buffer, which is
allocated in the memory range available to rtas 32-bit.
Reentrant rtas calls are useful to avoid deadlocks in crashing,
where rtas-calls are needed, but some other thread crashed holding
the rtas.lock.
This is a backtrace of a deadlock from a kdump testing environment:
#0 arch_spin_lock
#1 lock_rtas ()
#2 rtas_call (token=8204, nargs=1, nret=1, outputs=0x0)
#3 ics_rtas_mask_real_irq (hw_irq=4100)
#4 machine_kexec_mask_interrupts
#5 default_machine_crash_shutdown
#6 machine_crash_shutdown
#7 __crash_kexec
#8 crash_kexec
#9 oops_end
Signed-off-by: Leonardo Bras <leobras.c@gmail.com>
[mpe: Move under #ifdef PSERIES to avoid build breakage]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200518234245.200672-3-leobras.c@gmail.com
In order to get any rtas* struct into other headers, including rtas.h
may cause a lot of errors, regarding include dependency needed for
inline functions.
Create rtas-types.h and move there all type/struct definitions
from rtas.h, then include rtas-types.h into rtas.h.
Also, as suggested by checkpath.pl, replace uint8_t for u8.
Signed-off-by: Leonardo Bras <leobras.c@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200518234245.200672-2-leobras.c@gmail.com
Currently, if printk lock (logbuf_lock) is held by other thread during
crash, there is a chance of deadlocking the crash on next printk, and
blocking a possibly desired kdump.
At the start of default_machine_crash_shutdown, make printk enter
NMI context, as it will use per-cpu buffers to store the message,
and avoid locking logbuf_lock.
Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Leonardo Bras <leobras.c@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200512214533.93878-1-leobras.c@gmail.com
While providing guests, it's desirable to resize it's memory on demand.
By now, it's possible to do so by creating a guest with a small base
memory, hot-plugging all the rest, and using 'movable_node' kernel
command-line parameter, which puts all hot-plugged memory in
ZONE_MOVABLE, allowing it to be removed whenever needed.
But there is an issue regarding guest reboot:
If memory is hot-plugged, and then the guest is rebooted, all hot-plugged
memory goes to ZONE_NORMAL, which offers no guaranteed hot-removal.
It usually prevents this memory to be hot-removed from the guest.
It's possible to use device-tree information to fix that behavior, as
it stores flags for LMB ranges on ibm,dynamic-memory-vN.
It involves marking each memblock with the correct flags as hotpluggable
memory, which mm/memblock.c puts in ZONE_MOVABLE during boot if
'movable_node' is passed.
For carrying such information, the new flag DRCONF_MEM_HOTREMOVABLE was
proposed and accepted into Power Architecture documentation.
This flag should be:
- true (b=1) if the hypervisor may want to hot-remove it later, and
- false (b=0) if it does not care.
During boot, guest kernel reads the device-tree, early_init_drmem_lmb()
is called for every added LMBs. Here, checking for this new flag and
marking memblocks as hotplugable memory is enough to get the desirable
behavior.
This should cause no change if 'movable_node' parameter is not passed
in kernel command-line.
Signed-off-by: Leonardo Bras <leonardo@linux.ibm.com>
Reviewed-by: Bharata B Rao <bharata@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200402195156.626430-1-leonardo@linux.ibm.com
Show the address of the tasks regs in the process listing in xmon. The
regs should always be on the stack page that we also print the address
of, but it's still helpful not to have to find them by hand.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200520111740.953679-1-mpe@ellerman.id.au
This adds the CPU or thread number to printk messages. This helps a
lot when deciphering concurrent oopses that have been interleaved.
Example output, of PID1 (T1) triggering a warning:
[ 1.581678][ T1] WARNING: CPU: 0 PID: 1 at crypto/rsa-pkcs1pad.c:539 pkcs1pad_verify+0x38/0x140
[ 1.581681][ T1] Modules linked in:
[ 1.581693][ T1] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.5.0-rc5-gcc-8.2.0-00121-gf84c2e595927-dirty #1515
[ 1.581700][ T1] NIP: c000000000207d64 LR: c000000000207d3c CTR: c000000000207d2c
[ 1.581708][ T1] REGS: c0000000fd2e7560 TRAP: 0700 Not tainted (5.5.0-rc5-gcc-8.2.0-00121-gf84c2e595927-dirty)
[ 1.581712][ T1] MSR: 9000000000029033 <SF,HV,EE,ME,IR,DR,RI,LE> CR: 44000222 XER: 00040000
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200520121257.961112-1-mpe@ellerman.id.au
Currently when we boot on a big core system, we get this print:
[ 0.040500] Using small cores at SMT level
This is misleading as we've actually detected big cores.
This patch clears up the print to say we've detect big cores but are
using small cores for scheduling.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200528230731.1235752-1-mikey@neuling.org
If the memory chunk found for reserving memory overshoots the memory
limit imposed, do not proceed with reserving memory. Default behavior
was this until commit 140777a3d8 ("powerpc/fadump: consider reserved
ranges while reserving memory") changed it unwittingly.
Fixes: 140777a3d8 ("powerpc/fadump: consider reserved ranges while reserving memory")
Cc: stable@vger.kernel.org
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/159057266320.22331.6571453892066907320.stgit@hbathini.in.ibm.com
'mem=" option is an easy way to put high pressure on memory during
some test. Hence after applying the memory limit, instead of total
mem, the actual usable memory should be considered when reserving mem
for crashkernel. Otherwise the boot up may experience OOM issue.
E.g. it would reserve 4G prior to the change and 512M afterward, if
passing
crashkernel="2G-4G:384M,4G-16G:512M,16G-64G:1G,64G-128G:2G,128G-:4G",
and mem=5G on a 256G machine.
This issue is powerpc specific because it puts higher priority on
fadump and kdump reservation than on "mem=". Referring the following
code:
if (fadump_reserve_mem() == 0)
reserve_crashkernel();
...
/* Ensure that total memory size is page-aligned. */
limit = ALIGN(memory_limit ?: memblock_phys_mem_size(), PAGE_SIZE);
memblock_enforce_memory_limit(limit);
While on other arches, the effect of "mem=" takes a higher priority
and pass through memblock_phys_mem_size() before calling
reserve_crashkernel().
Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
Reviewed-by: Hari Bathini <hbathini@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1585749644-4148-1-git-send-email-kernelfans@gmail.com
kbuild test robot reported some build warnings in the hw_breakpoint
code when compiled with clang[1]. Some of them were introduced by the
recent powerpc change to add arch_reserve_bp_slot() and
arch_release_bp_slot(). Fix them all.
kernel/events/hw_breakpoint.c:71:12: warning: no previous prototype for function 'hw_breakpoint_weight'
kernel/events/hw_breakpoint.c:216:12: warning: no previous prototype for function 'arch_reserve_bp_slot'
kernel/events/hw_breakpoint.c:221:13: warning: no previous prototype for function 'arch_release_bp_slot'
kernel/events/hw_breakpoint.c:228:13: warning: no previous prototype for function 'arch_unregister_hw_breakpoint'
[1]: https://lore.kernel.org/linuxppc-dev/202005192233.oi9CjRtA%25lkp@intel.com/
Fixes: 29da4f91c0 ("powerpc/watchpoint: Don't allow concurrent perf and ptrace events")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
[mpe: Drop extern, flesh out change log, add Fixes tag]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200602041208.128913-1-ravi.bangoria@linux.ibm.com
XIVE interrupt controller uses an Event Queue (EQ) to enqueue event
notifications when an exception occurs. The EQ is a single memory page
provided by the O/S defining a circular buffer, one per server and
priority couple.
On baremetal, the EQ page is configured with an OPAL call. On pseries,
an extra hop is necessary and the guest OS uses the hcall
H_INT_SET_QUEUE_CONFIG to configure the XIVE interrupt controller.
The XIVE controller being Hypervisor privileged, it will not be allowed
to enqueue event notifications for a Secure VM unless the EQ pages are
shared by the Secure VM.
Hypervisor/Ultravisor still requires support for the TIMA and ESB page
fault handlers. Until this is complete, QEMU can use the emulated XIVE
device for Secure VMs, option "kernel_irqchip=off" on the QEMU pseries
machine.
Signed-off-by: Ram Pai <linuxram@us.ibm.com>
Reviewed-by: Cedric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200426020518.GC5853@oc0525413822.ibm.com
Function 'read_sys_info_pseries()' is added to get system parameter
values like number of sockets and chips per socket.
and it gets these details via rtas_call with token
"PROCESSOR_MODULE_INFO".
Incase lpar migrate from one system to another, system
parameter details like chips per sockets or number of sockets might
change. So, it needs to be re-initialized otherwise, these values
corresponds to previous system values.
This patch adds a call to 'read_sys_info_pseries()' from
'post-mobility_fixup()' to re-init the physsockets and physchips values
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200525104308.9814-6-kjain@linux.ibm.com
Add documentation for the following sysfs files:
/sys/devices/hv_24x7/interface/chipspersocket,
/sys/devices/hv_24x7/interface/sockets,
/sys/devices/hv_24x7/interface/coresperchip
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200525104308.9814-5-kjain@linux.ibm.com
To expose the system dependent parameter like total number of
sockets and numbers of chips per socket, patch adds two sysfs files.
"sockets" and "chips" are added to /sys/devices/hv_24x7/interface/
of the "hv_24x7" pmu.
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200525104308.9814-4-kjain@linux.ibm.com
For hv_24x7 socket/chip level events, specific chip-id to which
the data requested should be added as part of pmu events.
But number of chips/socket in the system details are not exposed.
Patch implements read_24x7_sys_info() to get system parameter values
like number of sockets, cores per chip and chips per socket. Rtas_call
with token "PROCESSOR_MODULE_INFO" is used to get these values.
Subsequent patch exports these values via sysfs.
Patch also make these parameters default to 1.
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200525104308.9814-3-kjain@linux.ibm.com
Commit 2b206ee6b0 ("powerpc/perf/hv-24x7: Display change in counter
values")' added to print _change_ in the counter value rather then raw
value for 24x7 counters. Incase of transactions, the event count
is set to 0 at the beginning of the transaction. It also sets
the event's prev_count to the raw value at the time of initialization.
Because of setting event count to 0, we are seeing some weird behaviour,
whenever we run multiple 24x7 events at a time.
For example:
command#: ./perf stat -e "{hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/,
hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/}"
-C 0 -I 1000 sleep 100
1.000121704 120 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
1.000121704 5 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
2.000357733 8 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
2.000357733 10 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
3.000495215 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
3.000495215 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
4.000641884 56 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
4.000641884 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
5.000791887 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
Getting these large values in case we do -I.
As we are setting event_count to 0, for interval case, overall event_count is not
coming in incremental order. As we may can get new delta lesser then previous count.
Because of which when we print intervals, we are getting negative value which create
these large values.
This patch removes part where we set event_count to 0 in function
'h_24x7_event_read'. There won't be much impact as we do set event->hw.prev_count
to the raw value at the time of initialization to print change value.
With this patch
In power9 platform
command#: ./perf stat -e "{hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/,
hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/}"
-C 0 -I 1000 sleep 100
1.000117685 93 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
1.000117685 1 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
2.000349331 98 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
2.000349331 2 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
3.000495900 131 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
3.000495900 4 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
4.000645920 204 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
4.000645920 61 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
4.284169997 22 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
Suggested-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Tested-by: Madhavan Srinivasan <maddy@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200525104308.9814-2-kjain@linux.ibm.com
pnv_pci_ioda_configure_bus() should now only ever be called when a device is
added to the bus so add a WARN_ON() to the empty bus check. Similarly,
pnv_pci_ioda_setup_bus_PE() should only ever be called for an unconfigured PE,
so add a WARN_ON() for that case too.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200417073508.30356-5-oohall@gmail.com
Doing it once during boot rather than doing it on the fly and drop the janky
populated logic.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200417073508.30356-4-oohall@gmail.com
For normal PHBs IODA PEs are handled on a per-bus basis so all the devices
on that bus will share a PE. Which PE specificly is determined by the location
of the MMIO BARs for the devices on the bus so we can't actually configure the
bus PEs until after MMIO resources are allocated. As a result PEs are currently
configured by pcibios_setup_bridge(), which is called just before the bridge
windows are programmed into the bus' parent bridge. Configuring the bus PE here
causes a few problems:
1. The root bus doesn't have a parent bridge so setting up the PE for the root
bus requires some hacks.
2. The PELT-V isn't setup correctly because pnv_ioda_set_peltv() assumes that
PEs will be configured in root-to-leaf order. This assumption is broken
because resource assignment is performed depth-first so the leaf bridges
are setup before their parents are. The hack mentioned in 1) results in
the "correct" PELT-V for busses immediately below the root port, but not
for devices below a switch.
3. It's possible to break the sysfs PCI rescan feature by removing all
the devices on a bus. When the last device is removed from a PE its
will be de-configured. Rescanning the devices on a bus does not cause
the bridge to be reconfigured rendering the devices on that bus
unusable.
We can address most of these problems by moving the PE setup out of
pcibios_setup_bridge() and into pcibios_bus_add_device(). This fixes 1)
and 2) because pcibios_bus_add_device() is called on each device in
root-to-leaf order so PEs for parent buses will always be configured
before their children. It also fixes 3) by ensuring the PE is
configured before initialising DMA for the device. In the event the PE
was de-configured due to removing all the devices in that PE it will
now be reconfigured when a new device is added since there's no
dependecy on the bridge_setup() hook being called.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200417073508.30356-3-oohall@gmail.com
For each PHB we maintain a reverse-map that can be used to find the
PE that a BDFN is currently mapped to. Add a helper for doing this
lookup so we can check if a PE has been configured without looking
at pdn->pe_number.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200417073508.30356-2-oohall@gmail.com
It's pretty obsecure and confused me for a long time so I figured it's
worth documenting properly.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200414233502.758-1-oohall@gmail.com
Quite useful to know in some cases.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Sam Bobroff <sbobroff@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200408112213.5549-1-oohall@gmail.com
The NVlink IOMMU group setup is only relevant to NVLink devices so move
it into the NPU containment zone. This let us remove some prototypes in
pci.h and staticfy some function definitions.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-8-oohall@gmail.com
Move it in with the rest of the TCE wrangling rather than carting around
a static prototype in pci-ioda.c
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-7-oohall@gmail.com
Historically adding devices to their respective iommu group has been
handled by the post-init phb fixup for most devices. This was done
because:
1) The IOMMU group is tied to the PE (usually) so we can only setup the
iommu groups after we've done resource allocation since BAR location
determines the device's PE, and:
2) The sysfs directory for the pci_dev needs to be available since
iommu_add_device() wants to add an attribute for the iommu group.
However, since commit 30d87ef8b3 ("powerpc/pci: Fix
pcibios_setup_device() ordering") both conditions are met when
hose->ops->dma_dev_setup() is called so there's no real need to do
this in the fixup.
Moving the call to iommu_add_device() into pnv_pci_ioda_dma_setup_dev()
is a nice cleanup since it puts all the per-device IOMMU setup into one
place. It also results in all (non-nvlink) devices getting their iommu
group via a common path rather than relying on the bus notifier hack
in pnv_tce_iommu_bus_notifier() to handle the adding VFs and
hotplugged devices to their group.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-5-oohall@gmail.com
Move the registration of IOMMU groups out of the post-phb init fixup and
into when we configure DMA for a PE. For most devices this doesn't
result in any functional changes, but for NVLink attached GPUs it
requires a bit of care. When the GPU is probed an IOMMU group would be
created for the PE that contains it. We need to ensure that group is
removed before we add the PE to the compound group that's used to keep
the translations see by the PCIe and NVLink buses the same.
No functional changes. Probably.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-4-oohall@gmail.com
In pnv_ioda_setup_vf_PE() we register an iommu group for the VF PE
then call pnv_ioda_setup_bus_iommu_group() to add devices to that group.
However, this function is called before the VFs are scanned so there's
no devices to add.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-3-oohall@gmail.com
Re-work the control flow a bit so what's going on is a little clearer.
This also ensures the table_group is only initialised once in the P9
case. This shouldn't be a functional change since all the GPU PCI
devices should have the same table_group configuration, but it does
look strange.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-2-oohall@gmail.com
Similar to the C code change, make the AMR restore conditional on
whether the register has changed.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429065654.1677541-7-npiggin@gmail.com
The AMR update is made conditional on AMR actually changing, which
should be the less common case on most workloads (though kernel page
faults on uaccess could be frequent, this doesn't significantly slow
down that case).
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429065654.1677541-4-npiggin@gmail.com
