A test kernel with the options DEBUG_TEST_DRIVER_REMOVE, KASAN and
DEBUG_KMEMLEAK set, revealed several issues when removing an mci device:
1) Use-after-free:
On 27.11.19 17:07:33, John Garry wrote:
> [ 22.104498] BUG: KASAN: use-after-free in
> edac_remove_sysfs_mci_device+0x148/0x180
The use-after-free is caused by the mci_for_each_dimm() macro called in
edac_remove_sysfs_mci_device(). The iterator was introduced with
c498afaf7d ("EDAC: Introduce an mci_for_each_dimm() iterator").
The iterator loop calls device_unregister(&dimm->dev), which removes
the sysfs entry of the device, but also frees the dimm struct in
dimm_attr_release(). When incrementing the loop in mci_for_each_dimm(),
the dimm struct is accessed again, after having been freed already.
The fix is to free all the mci device's subsequent dimm and csrow
objects at a later point, in _edac_mc_free(), when the mci device itself
is being freed.
This keeps the data structures intact and the mci device can be
fully used until its removal. The change allows the safe usage of
mci_for_each_dimm() to release dimm devices from sysfs.
2) Memory leaks:
Following memory leaks have been detected:
# grep edac /sys/kernel/debug/kmemleak | sort | uniq -c
1 [<000000003c0f58f9>] edac_mc_alloc+0x3bc/0x9d0 # mci->csrows
16 [<00000000bb932dc0>] edac_mc_alloc+0x49c/0x9d0 # csr->channels
16 [<00000000e2734dba>] edac_mc_alloc+0x518/0x9d0 # csr->channels[chn]
1 [<00000000eb040168>] edac_mc_alloc+0x5c8/0x9d0 # mci->dimms
34 [<00000000ef737c29>] ghes_edac_register+0x1c8/0x3f8 # see edac_mc_alloc()
All leaks are from memory allocated by edac_mc_alloc().
Note: The test above shows that edac_mc_alloc() was called here from
ghes_edac_register(), thus both functions show up in the stack trace
but the module causing the leaks is edac_mc. The comments with the data
structures involved were made manually by analyzing the objdump.
The data structures listed above and created by edac_mc_alloc() are
not properly removed during device removal, which is done in
edac_mc_free().
There are two paths implemented to remove the device depending on device
registration, _edac_mc_free() is called if the device is not registered
and edac_unregister_sysfs() otherwise.
The implemenations differ. For the sysfs case, the mci device removal
lacks the removal of subsequent data structures (csrows, channels,
dimms). This causes the memory leaks (see mci_attr_release()).
[ bp: Massage commit message. ]
Fixes: c498afaf7d ("EDAC: Introduce an mci_for_each_dimm() iterator")
Fixes: faa2ad09c0 ("edac_mc: edac_mc_free() cannot assume mem_ctl_info is registered in sysfs.")
Fixes: 7a623c0390 ("edac: rewrite the sysfs code to use struct device")
Reported-by: John Garry <john.garry@huawei.com>
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Tested-by: John Garry <john.garry@huawei.com>
Cc: <stable@vger.kernel.org>
Link: https://lkml.kernel.org/r/20200212120340.4764-3-rrichter@marvell.com
The code in ghes_edac.c and edac_mc.c for grain_bits calculation and
calling trace_mc_event() is now the same. Move it to a single location
in edac_raw_mc_handle_error().
The only difference is the missing IS_ENABLED(CONFIG_RAS) switch, but
this is needed for ghes too.
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20191106093239.25517-13-rrichter@marvell.com
The e string to which this is pointing to has already been cleared
earlier in the function so remove the needless zero string termination.
[ bp: Correct the commit message. ]
Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20191106093239.25517-6-rrichter@marvell.com
No need to crash the system in case edac_mc_alloc() is called with
invalid arguments, just warn and return. This would cause a checkpatch
warning when touching the code later, so just fix it.
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20191106093239.25517-5-rrichter@marvell.com
Introduce an mci_for_each_dimm() iterator. It returns a pointer to
a struct dimm_info. This makes the declaration and use of an index
obsolete and avoids access to internal data of struct mci (direct array
access etc).
[ bp: push the struct dimm_info *dimm; declaration into the
CONFIG_EDAC_DEBUG block. ]
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20191106093239.25517-4-rrichter@marvell.com
The EDAC_DIMM_OFF() macro takes 5 arguments to get the DIMM's index.
Simplify this by storing the index in struct dimm_info to avoid its
calculation and remove the EDAC_DIMM_OFF() macro. The index can be
directly used then.
Another advantage is that edac_mc_alloc() could be used even if the
exact size of the layers is unknown. Only the number of DIMMs would be
needed.
Rename iterator variable to idx, while at it. The name is more handy,
esp. when searching for it in the code.
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20191106093239.25517-3-rrichter@marvell.com
Use of 'unsigned int' instead of bare use of 'unsigned'. Fix this for
edac_mc*, ghes and the i5100 driver as reported by checkpatch.pl.
While at it, struct member dev_ch_attribute->channel is always used as
unsigned int. Change type to unsigned int to avoid type casts.
[ bp: Massage. ]
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20190902123216.9809-2-rrichter@marvell.com
The grain in EDAC is defined as "minimum granularity for an error
report, in bytes". The following calculation of the grain_bits in
edac_mc is wrong:
grain_bits = fls_long(e->grain) + 1;
Where grain_bits is defined as:
grain = 1 << grain_bits
Example:
grain = 8 # 64 bit (8 bytes)
grain_bits = fls_long(8) + 1
grain_bits = 4 + 1 = 5
grain = 1 << grain_bits
grain = 1 << 5 = 32
Replace it with the correct calculation:
grain_bits = fls_long(e->grain - 1);
The example gives now:
grain_bits = fls_long(8 - 1)
grain_bits = fls_long(7)
grain_bits = 3
grain = 1 << 3 = 8
Also, check if the hardware reports a reasonable grain != 0 and fallback
with a warning to 1 byte granularity otherwise.
[ bp: massage a bit. ]
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20190624150758.6695-2-rrichter@marvell.com
The function should return NULL in case no device is found, but it
always returns the last checked mc device from the list even if the
index did not match. Fix that.
I did some analysis why this did not raise any issues for about 3 years
and the reason is that edac_mc_find() is mostly used to search for
existing devices. Thus, the bug is not triggered.
[ bp: Drop the if (mci->mc_idx > idx) test in favor of readability. ]
Fixes: c73e8833be ("EDAC, mc: Fix locking around mc_devices list")
Signed-off-by: Robert Richter <rrichter@marvell.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: "linux-edac@vger.kernel.org" <linux-edac@vger.kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Link: https://lkml.kernel.org/r/20190514104838.15065-1-rrichter@marvell.com
... and use the single edac_subsys object returned from
subsys_system_register(). The idea is to have a single bus
and multiple devices on it.
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
CC: Aristeu Rozanski Filho <arozansk@redhat.com>
CC: Greg KH <gregkh@linuxfoundation.org>
CC: Justin Ernst <justin.ernst@hpe.com>
CC: linux-edac <linux-edac@vger.kernel.org>
CC: Mauro Carvalho Chehab <mchehab@kernel.org>
CC: Russ Anderson <rja@hpe.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20180926152752.GG5584@zn.tnic
There are now non-volatile versions of DIMMs. Add a new entry to "enum
mem_type" and a new string in edac_mem_types[].
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Aristeu Rozanski <aris@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Jean Delvare <jdelvare@suse.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: linux-acpi@vger.kernel.org
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: linux-nvdimm@lists.01.org
Link: http://lkml.kernel.org/r/20180312182430.10335-3-tony.luck@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Somehow we ended up with two separate arrays of strings to describe the
"enum mem_type" values.
In edac_mc.c we have an exported list edac_mem_types[] that is used
by a couple of drivers in debug messaged.
In edac_mc_sysfs.c we have a private list that is used to display
values in:
/sys/devices/system/edac/mc/mc*/dimm*/dimm_mem_type
/sys/devices/system/edac/mc/mc*/csrow*/mem_type
This list was missing a value for MEM_LRDDR3.
The string values in the two lists were different :-(
Combining the lists, I kept the values so that the sysfs output
will be unchanged as some scripts may depend on that.
Reported-by: Borislav Petkov <bp@suse.de>
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Aristeu Rozanski <aris@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Jean Delvare <jdelvare@suse.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: linux-acpi@vger.kernel.org
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: linux-nvdimm@lists.01.org
Link: http://lkml.kernel.org/r/20180312182430.10335-2-tony.luck@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Only a single EDAC platform driver can be loaded. When ghes_edac is
enabled, an EDAC platform driver still attempts to register itself and
fails in edac_mc_add_mc().
Add edac_get_owner() so that EDAC platform drivers can check the owner
first.
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Suggested-by: Borislav Petkov <bp@alien8.de>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170823225447.15608-5-toshi.kani@hpe.com
[ Massage commit message. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
Use mc_devices list instead to check whether we have EDAC driver
instances successfully registered with EDAC core.
Signed-off-by: Borislav Petkov <bp@suse.de>
This was entirely automated, using the script by Al:
PATT='^[[:blank:]]*#[[:blank:]]*include[[:blank:]]*<asm/uaccess.h>'
sed -i -e "s!$PATT!#include <linux/uaccess.h>!" \
$(git grep -l "$PATT"|grep -v ^include/linux/uaccess.h)
to do the replacement at the end of the merge window.
Requested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Several functions are documented at edac_mc.c.
As we'll be including edac_core.h at drivers-api book, move
those, in order for the kernel-doc markups be part of the API
documentation book.
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
When accessing the mc_devices list of memory controller descriptors, we
need to hold mem_ctls_mutex. This was not always the case, fix that.
Make all external callers call a version which grabs the mutex since the
last is local to edac_mc.c.
Reported-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
They're both running only when ->edac_check is initialized so remove
that check from the workqueue function itself. Synchronize/generalize
the ->op_state check between the two.
Kill useless comments, while at it.
Signed-off-by: Borislav Petkov <bp@suse.de>
We use the ->edac_check function pointers to determine whether we need
to setup a polling workqueue. However, the destroy path is not balanced
and we might try to teardown an unitialized workqueue.
Balance init and destroy paths by looking at ->edac_check in both cases.
Set op_state to OP_OFFLINE *before* destroying anything.
Reported-by: Zhiqiang Hou <Zhiqiang.Hou@freescale.com>
Cc: Varun Sethi <Varun.Sethi@freescale.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Hide the EDAC workqueue pointer in a separate compilation unit and add
accessors for the workqueue manipulations needed.
Remove edac_pci_reset_delay_period() which wasn't used by anything. It
seems it got added without a user with
91b99041c1 ("drivers/edac: updated PCI monitoring")
Signed-off-by: Borislav Petkov <bp@suse.de>
EDAC workqueue destruction is really fragile. We cancel delayed work
but if it is still running and requeues itself, we still go ahead and
destroy the workqueue and the queued work explodes when workqueue core
attempts to run it.
Make the destruction more robust by switching op_state to offline so
that requeuing stops. Cancel any pending work *synchronously* too.
EDAC i7core: Driver loaded.
general protection fault: 0000 [#1] SMP
CPU 12
Modules linked in:
Supported: Yes
Pid: 0, comm: kworker/0:1 Tainted: G IE 3.0.101-0-default #1 HP ProLiant DL380 G7
RIP: 0010:[<ffffffff8107dcd7>] [<ffffffff8107dcd7>] __queue_work+0x17/0x3f0
< ... regs ...>
Process kworker/0:1 (pid: 0, threadinfo ffff88019def6000, task ffff88019def4600)
Stack:
...
Call Trace:
call_timer_fn
run_timer_softirq
__do_softirq
call_softirq
do_softirq
irq_exit
smp_apic_timer_interrupt
apic_timer_interrupt
intel_idle
cpuidle_idle_call
cpu_idle
Code: ...
RIP __queue_work
RSP <...>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: <stable@vger.kernel.org>
The PAGES_TO_MiB macro is used for unit conversion but the
trace_mc_event() tracepoint expects a page address. Fix that.
Signed-off-by: Tan Xiaojun <tanxiaojun@huawei.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/1445341538-24271-1-git-send-email-tanxiaojun@huawei.com
Signed-off-by: Borislav Petkov <bp@suse.de>
So first of all, this atomic_scrub() function's naming is bad. It looks
like an atomic_t helper. Change it to edac_atomic_scrub().
The bigger problem is that this function is arch-specific and every new
arch which doesn't necessarily need that functionality still needs to
define it, otherwise EDAC doesn't compile.
So instead of doing that and including arch-specific headers, have each
arch define an EDAC_ATOMIC_SCRUB symbol which can be used in edac_mc.c
for ifdeffery. Much cleaner.
And we already are doing this with another symbol - EDAC_SUPPORT. This
is also much cleaner than having CONFIG_EDAC enumerate all the arches
which need/have EDAC support and drivers.
This way I can kill the useless edac.h header in tile too.
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Acked-by: Chris Metcalf <cmetcalf@ezchip.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Doug Thompson <dougthompson@xmission.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-edac@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: linux-mips@linux-mips.org
Cc: linuxppc-dev@lists.ozlabs.org
Cc: "Maciej W. Rozycki" <macro@codesourcery.com>
Cc: Markos Chandras <markos.chandras@imgtec.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Steven J. Hill" <Steven.Hill@imgtec.com>
Cc: x86@kernel.org
Signed-off-by: Borislav Petkov <bp@suse.de>
Add edac_mc_add_mc_with_groups() for initializing the mem_ctl_info
object with the optional attribute groups. This allows drivers to
pass additional sysfs entries without manual (and racy)
device_create_file() and co calls.
edac_mc_add_mc() is kept as is, just calling edac_mc_add_with_groups()
with NULL groups.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: http://lkml.kernel.org/r/1423046938-18111-3-git-send-email-tiwai@suse.de
Signed-off-by: Borislav Petkov <bp@suse.de>
Make keeping the sync between the mem_types enum and the actual string
names simpler by using designated initializers.
Signed-off-by: Borislav Petkov <bp@suse.de>
To avoid confuision and conflict of usage for RAS related trace event,
add an unified RAS trace event stub.
Start a RAS subsystem menu which will be fleshed out in time, when more
features get added to it.
Signed-off-by: Chen, Gong <gong.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/1402475691-30045-2-git-send-email-gong.chen@linux.intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
We're using edac_mc_workq_setup() both on the init path, when
we load an edac driver and when we change the polling period
(edac_mc_reset_delay_period) through /sys/.../edac_mc_poll_msec.
On that second path we don't need to init the workqueue which has been
initialized already.
Thanks to Tejun for workqueue insights.
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: http://lkml.kernel.org/r/1391457913-881-1-git-send-email-prarit@redhat.com
Cc: <stable@vger.kernel.org>
Sanitize code even more to accept unsigned longs only and to not allow
polling intervals below 1 second as this is unnecessary and doesn't make
much sense anyway for polling errors.
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: http://lkml.kernel.org/r/1391457913-881-1-git-send-email-prarit@redhat.com
Cc: Doug Thompson <dougthompson@xmission.com>
Cc: <stable@vger.kernel.org>
Log messages slightly differ between edac subsystems. Unifying it.
Signed-off-by: Robert Richter <robert.richter@linaro.org>
Acked-by: Rob Herring <rob.herring@calxeda.com>
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Robert Richter <rric@kernel.org>
Both mci.mem_is_per_rank and mci.csbased denote the same thing: the
memory controller is csrows based. Merge both fields into one.
There's no need for the driver to actually fill it, as the core detects
it by checking if one of the layers has the csrows type as part of the
memory hierarchy:
if (layers[i].type == EDAC_MC_LAYER_CHIP_SELECT)
per_rank = true;
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
The number of variables at the stack is too big.
Reduces the stack usage by using a pre-allocated error
buffer.
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
APEI GHES and i7core_edac/sb_edac currently can be loaded at
the same time, but those are Highlander modules:
"There can be only one".
There are two reasons for that:
1) Each driver assumes that it is the only one registering at
the EDAC core, as it is driver's responsibility to number
the memory controllers, and all of them start from 0;
2) If BIOS is handling the memory errors, the OS can't also be
doing it, as one will mangle with the other.
So, we need to add an module owner's lock at the EDAC core,
in order to avoid having two different modules handling memory
errors at the same time. The best way for doing this lock seems
to use the driver's name, as this is unique, and won't require
changes on every driver.
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
