To reduce the size of builtin-trace.c.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-vb8dpy7bptkf219q5c25ulfp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To reduce the size of builtin-trace.c.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-jt293541hv9od7gqw6lilioh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To reduce the size of builtin-trace.c.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-qecqxwwtreio6eaatfv58yq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add debug output of raw counter values per CPU when perf stat -v is
specified, together with their cpu numbers. This is very useful to
debug problems with per core counters, where we can normally only see
aggregated values.
v2: Make it depend on -vv, not -v
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1461787251-6702-12-git-send-email-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Update the export-to-postgresql.py to support the newly introduced
callchain export.
callchains are added into the existing call_paths table and can now
be associated with samples when the "callpaths" commandline option
is used with the script.
Ex.:
$ perf script -s export-to-postgresql.py example_db all callchains
Includes the following changes to enable callchain export via the python export
APIs:
- Add the "callchains" commandline option, which is used to enable
callchain export by setting the perf_db_export_callchains global
- Add perf_db_export_callchains checks for call_path table creation
and population.
- Add call_path_id to samples_table to conform with the new API
example usage and output using a small test app:
test_app.c:
volatile int x = 0;
void inc_x_loop()
{
int i;
for(i=0; i<100000000; i++)
x++;
}
void a()
{
inc_x_loop();
}
void b()
{
inc_x_loop();
}
int main()
{
a();
b();
return 0;
}
example usage:
$ gcc -g -O0 test_app.c
$ perf record --call-graph=dwarf ./a.out
[ perf record: Woken up 77 times to write data ]
[ perf record: Captured and wrote 19.373 MB perf.data (2404 samples) ]
$ perf script -s scripts/python/export-to-postgresql.py
example_db all callchains
$ psql example_db
example_db=#
SELECT
(SELECT name FROM symbols WHERE id = cps.symbol_id) as symbol,
(SELECT name FROM symbols WHERE id =
(SELECT symbol_id from call_paths where id = cps.parent_id))
as parent_symbol,
sum(period) as event_count
FROM samples join call_paths as cps on call_path_id = cps.id
GROUP BY cps.id,evsel_id
ORDER BY event_count DESC
LIMIT 5;
symbol | parent_symbol | event_count
------------------+--------------------------+-------------
inc_x_loop | a | 734250982
inc_x_loop | b | 731028057
unknown | unknown | 1335858
task_tick_fair | scheduler_tick | 1238842
update_wall_time | tick_do_update_jiffies64 | 650373
(5 rows)
The above data shows total "self time" in cycles for each call path that was
sampled. It is intended to demonstrate how it accounts separately for the two
ways to reach the "inc_x_loop" function(via "a" and "b"). Recursive common
table expressions can be used as well to get cumulative time spent in a
function as well, but that is beyond the scope of this basic example.
Signed-off-by: Chris Phlipot <cphlipot0@gmail.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1461831551-12213-7-git-send-email-cphlipot0@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This change allows python scripts to be able to utilize the recent
changes to the db export api allowing the export of call_paths derived
from sampled callchains. These call paths are also now associated with
the samples from which they were derived.
- This feature is enabled by setting "perf_db_export_callchains" to true
- When enabled, samples that have callchain information will have the
callchains exported via call_path_table
- The call_path_id field is added to sample_table to enable association of
samples with the corresponding callchain stored in the call paths
table. A call_path_id of 0 will be exported if there is no
corresponding callchain.
- When "perf_db_export_callchains" and "perf_db_export_calls" are both
set to True, the call path root data structure will be shared. This
prevents duplicating of data and call path ids that would result from
building two separate call path trees in memory.
- The call_return_processor structure definition was relocated to the header
file to make its contents visible to db-export.c. This enables the
sharing of call path trees between the two features, as mentioned
above.
This change is visible to python scripts using the python db export api.
The change is backwards compatible with scripts written against the
previous API, assuming that the scripts model the sample_table function
after the one in export-to-postgresql.py script by allowing for
additional arguments to be added in the future. ie. using *x as the
final argument of the sample_table function.
Signed-off-by: Chris Phlipot <cphlipot0@gmail.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1461831551-12213-6-git-send-email-cphlipot0@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The exported sample now contains a reference to the call_path_id that
represents its callchain.
While callchains themselves are nice to have, being able to associate
them with samples makes them much more useful, and can allow for such
things as determining how much cumulative time is spent in a particular
function. This information is normally possible to get from the call
return processor. However, when doing normal sampling, call/return
information is not available, thus necessitating the need for
associating samples directly with call paths.
This commit include changes to db-export layer to make this information
available for subsequent patches in this change set, but by itself, does
not make any changes visible to the user.
Signed-off-by: Chris Phlipot <cphlipot0@gmail.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1461831551-12213-5-git-send-email-cphlipot0@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This change enables the db export api to export callchains. This is
accomplished by adding callchains obtained from samples to the
call_path_root structure and exporting them via the current call path
export API.
While the current API does support exporting call paths, this is not
supported when sampling. This commit addresses that missing feature by
allowing the export of call paths when callchains are present in
samples.
Summary:
- This feature is activated by initializing the call_path_root member
inside the db_export structure to a non-null value.
- Callchains are resolved with thread__resolve_callchain() and then stored
and exported by adding a call path under call path root.
- Symbol and DSO for each callchain node are exported via db_ids_from_al()
This commit puts in place infrastructure to be used by subsequent commits,
and by itself, does not introduce any user-visible changes.
Signed-off-by: Chris Phlipot <cphlipot0@gmail.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1461831551-12213-4-git-send-email-cphlipot0@gmail.com
[ Made adjustments suggested by Adrian Hunter, see thread via this cset's Link: tag ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Move the call path handling code out of thread-stack.c and
thread-stack.h to allow other components that are not part of
thread-stack to create call paths.
Summary:
- Create call-path.c and call-path.h and add them to the build.
- Move all call path related code out of thread-stack.c and thread-stack.h
and into call-path.c and call-path.h.
- A small subset of structures and functions are now visible through
call-path.h, which is required for thread-stack.c to continue to
compile.
This change is a prerequisite for subsequent patches in this change set
and by itself contains no user-visible changes.
Signed-off-by: Chris Phlipot <cphlipot0@gmail.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1461831551-12213-3-git-send-email-cphlipot0@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Do not load one entry beyond the end of the syscall table when the
syscall number of a traced process equals to __NR_Linux_syscalls.
Similar bug with regular processes was fixed by commit 3bb457af4f
("[PARISC] Fix bug when syscall nr is __NR_Linux_syscalls").
This bug was found by strace test suite.
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Acked-by: Helge Deller <deller@gmx.de>
Signed-off-by: Helge Deller <deller@gmx.de>
No need for an extra notifier. We don't need to handle all these states. It's
sufficient to kill the timer when the cpu dies.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160310120025.770528462@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The scheduler can handle per cpu threads before the cpu is set to active and
it does not allow user space threads on the cpu before active is
set. Attaching to the scheduling domains is also not required before user
space threads can be handled.
Move the activation to the end of the hotplug state space. That also means
that deactivation is the first action when a cpu is shut down.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160310120025.597477199@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The alleged requirement that the migration notifier has a lower priority than
perf is completely undocumented and there is no indication at all that this is
true. perf does not even handle the CPU_ONLINE notification and perf really
has nothing to do with migration.
Move the CPU_ONLINE code into the sched_activate_cpu() state callback.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160310120025.421743581@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
It really does not matter when we fold the load for the outgoing cpu. It's
almost dead anyway, so there is no harm if we fail to fold the few
microseconds which are required for going fully away.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160310120025.328739226@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
We can piggy pack that on the SCHED_STARTING state. It's not required before
the cpu actually comes online. Name the function proper as it has nothing to
do with migration.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160310120025.248226511@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The sync_rcu stuff is specificically for clearing bits in the active
mask, such that everybody will observe the bit cleared and will not
consider the cleared CPU for load-balancing etc.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160310120025.169219710@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Now that we reduced everything into single notifiers, it's simple to move them
into the hotplug state machine space.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This is the last operation on the cpu before vanishing. No point in calling
that on CPU_DEAD.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
We can maintain the ordering of the scheduler cpu hotplug functionality nicely
in one notifer. Get rid of the maze.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Prevent the SMP scheduler related notifiers to be executed before the smp
scheduler is initialized and install them early.
This is a preparatory change for further consolidation of the hotplug notifier
maze.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Start distangling the maze of hotplug notifiers in the scheduler.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: rt@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
In order to enable symmetric hotplug, we must mirror the online &&
!active state of cpu-down on the cpu-up side.
However, to retain sanity, limit this state to per-cpu kthreads.
Aside from the change to set_cpus_allowed_ptr(), which allow moving
the per-cpu kthreads on, the other critical piece is the cpu selection
for pinned tasks in select_task_rq(). This avoids dropping into
select_fallback_rq().
select_fallback_rq() cannot be allowed to select !active cpus because
its used to migrate user tasks away. And we do not want to move user
tasks onto cpus that are in transition.
Requested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: Jan H. Schönherr <jschoenh@amazon.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160301152303.GV6356@twins.programming.kicks-ass.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The existing implementation of thread__resolve_callchain, under certain
circumstances, can assemble callchain entries in the incorrect order.
The callchain entries are resolved incorrectly for a sample when all of
the following conditions are met:
1. callchain_param.order is set to ORDER_CALLER
2. thread__resolve_callchain_sample is able to resolve callchain entries
for the sample.
3. unwind__get_entries is also able to resolve callchain entries for the
sample.
The fix is accomplished by reversing the order in which
thread__resolve_callchain_sample and unwind__get_entries are called when
callchain_param.order is set to ORDER_CALLER.
Unwind specific code from thread__resolve_callchain is also moved into a
new static function to improve readability of the fix.
How to Reproduce the Existing Bug:
Modifying perf script to print call trees in the opposite order or
applying the remaining patches from this series and comparing the
results output from export-to-postgtresql.py are the easiest ways to see
the bug, however it can still be seen in current builds using perf
report.
Here is how i can reproduce the bug using perf report:
# perf record --call-graph=dwarf stress -c 1 -t 5
when i run this command:
# perf report --call-graph=flat,0,0,callee
This callchain, containing kernel (handle_irq_event, etc) and userspace
samples (__libc_start_main, etc) is contained in the output, which looks
correct (callee order):
gen8_irq_handler
handle_irq_event_percpu
handle_irq_event
handle_edge_irq
handle_irq
do_IRQ
ret_from_intr
__random
rand
0x558f2a04dded
0x558f2a04c774
__libc_start_main
0x558f2a04dcd9
Now run this command using caller order:
# perf report --call-graph=flat,0,0,caller
It is expected to see the exact reverse of the above when using caller
order (with "0x558f2a04dcd9" at the top and "gen8_irq_handler" at the
bottom) in the output, but it is nowhere to be found.
instead you see this:
ret_from_intr
do_IRQ
handle_irq
handle_edge_irq
handle_irq_event
handle_irq_event_percpu
gen8_irq_handler
0x558f2a04dcd9
__libc_start_main
0x558f2a04c774
0x558f2a04dded
rand
__random
Notice how internally the kernel symbols are reversed and the user space
symbols are reversed, but the kernel symbols still appear above the user
space symbols.
if this patch is applied and perf script is re-run, you will see the
expected output (with "0x558f2a04dcd9" at the top and "gen8_irq_handler"
at the bottom):
0x558f2a04dcd9
__libc_start_main
0x558f2a04c774
0x558f2a04dded
rand
__random
ret_from_intr
do_IRQ
handle_irq
handle_edge_irq
handle_irq_event
handle_irq_event_percpu
gen8_irq_handler
Signed-off-by: Chris Phlipot <cphlipot0@gmail.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1461831551-12213-2-git-send-email-cphlipot0@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The test to check if the arg format had been read from the
syscall:sys_enter_name/format file was looking at the list of non-commom
fields, and if that is empty, it would think it had failed to read it,
because it doesn't exist, for instance, for the clone() syscall.
So instead before dumping the raw syscall args list check
IS_ERR(sc->tp_format), if that is true, then an attempt was made to read
the format file and failed, in which case dump the raw arg list values.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-ls7pmdqb2xy9339vdburwvnk@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
* pm-opp-fixes:
PM / OPP: Remove useless check
* pm-cpufreq-fixes:
intel_pstate: Fix intel_pstate_get()
cpufreq: intel_pstate: Fix HWP on boot CPU after system resume
cpufreq: st: enable selective initialization based on the platform
* pm-cpuidle-fixes:
ARM: cpuidle: Pass on arm_cpuidle_suspend()'s return value
Currently we read the tsc radio: ratio = (MSR_PLATFORM_INFO >> 8) & 0x1f;
Thus we get bit 8-12 of MSR_PLATFORM_INFO, however according to the SDM
(35.5), the ratio bits are bit 8-15.
Ignoring the upper bits can result in an incorrect tsc ratio, which causes the
TSC calibration and the Local APIC timer frequency to be incorrect.
Fix this problem by masking 0xff instead.
[ tglx: Massaged changelog ]
Fixes: 7da7c15613 "x86, tsc: Add static (MSR) TSC calibration on Intel Atom SoCs"
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: stable@vger.kernel.org
Cc: Bin Gao <bin.gao@intel.com>
Cc: Len Brown <lenb@kernel.org>
Link: http://lkml.kernel.org/r/1462505619-5516-1-git-send-email-yu.c.chen@intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The mem_avoid[] array is used to track positions that should be avoided (like
the compressed kernel, decompression code, etc) when selecting a memory
position for the randomly relocated kernel. Since ZO is now at the end of
the decompression buffer and the decompression code (and its heap and
stack) are at the front, we can safely consolidate the decompression entry,
the heap entry, and the stack entry. The boot_params memory, however, could
be elsewhere, so it should be explicitly included.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Baoquan He <bhe@redhat.com>
[ Rwrote changelog, cleaned up code comments. ]
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: kernel-hardening@lists.openwall.com
Cc: lasse.collin@tukaani.org
Link: http://lkml.kernel.org/r/1462486436-3707-3-git-send-email-keescook@chromium.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Currently extract_kernel() defines the input and output buffer pointers
as "unsigned char *" since that's effectively what they are. It passes
these to the decompressor routine and to the ELF parser, which both
logically deal with buffer pointers too. There is some casting ("unsigned
long") done to validate the numerical value of the pointers, but it is
relatively limited.
However, choose_random_location() operates almost exclusively on the
numerical representation of these pointers, so it ended up carrying
a lot of "unsigned long" casts. With the future physical/virtual split
these casts were going to multiply, so this attempts to solve the
problem by doing all the casting in choose_random_location()'s entry
and return instead of through-out the code. Adjusts argument names to
be more meaningful, and changes one us of "choice" to "output" to make
the future physical/virtual split more clear (i.e. "choice" should be
strictly a function return value and not used as an intermediate).
Suggested-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Baoquan He <bhe@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: kernel-hardening@lists.openwall.com
Cc: lasse.collin@tukaani.org
Link: http://lkml.kernel.org/r/1462486436-3707-2-git-send-email-keescook@chromium.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
User visible:
- Order output of 'perf trace --summary' better, now the threads will
appear ascending order of number of events, and then, for each, in
descending order of syscalls by the time spent in the syscalls, so
that the last page produced can be the one about the most interesting
thread straced, suggested by Milian Wolff (Arnaldo Carvalho de Melo)
- Do not show the runtime_ms for a thread when not collecting it, that
is done so far only with 'perf trace --sched' (Arnaldo Carvalho de Melo)
- Fix kallsyms perf test on ppc64le (Naveen N. Rao)
Infrastructure:
- Move global variables related to presence of some keys in the sort order to a
per hist struct, to allow code like the hists browser to work with multiple
hists with different lists of columns (Jiri Olsa)
- Add support for generating bpf prologue in powerpc (Naveen N. Rao)
- Fix kprobe and kretprobe handling with kallsyms on ppc64le (Naveen N. Rao)
- evlist mmap changes, prep work for supporting reading backwards (Wang Nan)
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=zIqy
-----END PGP SIGNATURE-----
Merge tag 'perf-core-for-mingo-20160505' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core
Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:
User visible changes:
- Order output of 'perf trace --summary' better, now the threads will
appear ascending order of number of events, and then, for each, in
descending order of syscalls by the time spent in the syscalls, so
that the last page produced can be the one about the most interesting
thread straced, suggested by Milian Wolff (Arnaldo Carvalho de Melo)
- Do not show the runtime_ms for a thread when not collecting it, that
is done so far only with 'perf trace --sched' (Arnaldo Carvalho de Melo)
- Fix kallsyms perf test on ppc64le (Naveen N. Rao)
Infrastructure changes:
- Move global variables related to presence of some keys in the sort order to a
per hist struct, to allow code like the hists browser to work with multiple
hists with different lists of columns (Jiri Olsa)
- Add support for generating bpf prologue in powerpc (Naveen N. Rao)
- Fix kprobe and kretprobe handling with kallsyms on ppc64le (Naveen N. Rao)
- evlist mmap changes, prep work for supporting reading backwards (Wang Nan)
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Merge fixes from Andrew Morton:
"14 fixes"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
byteswap: try to avoid __builtin_constant_p gcc bug
lib/stackdepot: avoid to return 0 handle
mm: fix kcompactd hang during memory offlining
modpost: fix module autoloading for OF devices with generic compatible property
proc: prevent accessing /proc/<PID>/environ until it's ready
mm/zswap: provide unique zpool name
mm: thp: kvm: fix memory corruption in KVM with THP enabled
MAINTAINERS: fix Rajendra Nayak's address
mm, cma: prevent nr_isolated_* counters from going negative
mm: update min_free_kbytes from khugepaged after core initialization
huge pagecache: mmap_sem is unlocked when truncation splits pmd
rapidio/mport_cdev: fix uapi type definitions
mm: memcontrol: let v2 cgroups follow changes in system swappiness
mm: thp: correct split_huge_pages file permission
get_bridge_ifindices() is used from the old "deviceless" bridge ioctl
calls which aren't called with rtnl held. The comment above says that it is
called with rtnl but that is not really the case.
Here's a sample output from a test ASSERT_RTNL() which I put in
get_bridge_ifindices and executed "brctl show":
[ 957.422726] RTNL: assertion failed at net/bridge//br_ioctl.c (30)
[ 957.422925] CPU: 0 PID: 1862 Comm: brctl Tainted: G W O
4.6.0-rc4+ #157
[ 957.423009] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS 1.8.1-20150318_183358- 04/01/2014
[ 957.423009] 0000000000000000 ffff880058adfdf0 ffffffff8138dec5
0000000000000400
[ 957.423009] ffffffff81ce8380 ffff880058adfe58 ffffffffa05ead32
0000000000000001
[ 957.423009] 00007ffec1a444b0 0000000000000400 ffff880053c19130
0000000000008940
[ 957.423009] Call Trace:
[ 957.423009] [<ffffffff8138dec5>] dump_stack+0x85/0xc0
[ 957.423009] [<ffffffffa05ead32>]
br_ioctl_deviceless_stub+0x212/0x2e0 [bridge]
[ 957.423009] [<ffffffff81515beb>] sock_ioctl+0x22b/0x290
[ 957.423009] [<ffffffff8126ba75>] do_vfs_ioctl+0x95/0x700
[ 957.423009] [<ffffffff8126c159>] SyS_ioctl+0x79/0x90
[ 957.423009] [<ffffffff8163a4c0>] entry_SYSCALL_64_fastpath+0x23/0xc1
Since it only reads bridge ifindices, we can use rcu to safely walk the net
device list. Also remove the wrong rtnl comment above.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The peer may be expecting a reply having sent a request and then done a
shutdown(SHUT_WR), so tearing down the whole socket at this point seems
wrong and breaks for me with a client which does a SHUT_WR.
Looking at other socket family's stream_recvmsg callbacks doing a shutdown
here does not seem to be the norm and removing it does not seem to have
had any adverse effects that I can see.
I'm using Stefan's RFC virtio transport patches, I'm unsure of the impact
on the vmci transport.
Signed-off-by: Ian Campbell <ian.campbell@docker.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>
Cc: Andy King <acking@vmware.com>
Cc: Dmitry Torokhov <dtor@vmware.com>
Cc: Jorgen Hansen <jhansen@vmware.com>
Cc: Adit Ranadive <aditr@vmware.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Use htons instead of unconditionally byte swapping nexthdr. On a little
endian systems shifting the byte is correct behavior, but it results in
incorrect csums on big endian architectures.
Fixes: f8c6455bb0 ('net/mlx4_en: Extend checksum offloading by CHECKSUM COMPLETE')
Signed-off-by: Daniel Jurgens <danielj@mellanox.com>
Reviewed-by: Carol Soto <clsoto@us.ibm.com>
Tested-by: Carol Soto <clsoto@us.ibm.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull libnvdimm fixes from Dan Williams:
- a fix for the persistent memory 'struct page' driver. The
implementation overlooked the fact that pages are allocated in 2MB
units leading to -ENOMEM when establishing some configurations.
It's tagged for -stable as the problem was introduced with the
initial implementation in 4.5.
- The new "error status translation" routine, introduced with the 4.6
updates to the nfit driver, missed a necessary path in
acpi_nfit_ctl().
The end result is that we are falsely assuming commands complete
successfully when the embedded status says otherwise.
* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
nfit: fix translation of command status results
libnvdimm, pfn: fix memmap reservation sizing
This is another attempt to avoid a regression in wwn_to_u64() after that
started using get_unaligned_be64(), which in turn ran into a bug on
gcc-4.9 through 6.1.
The regression got introduced due to the combination of two separate
workarounds (commits e3bde9568d: "include/linux/unaligned: force
inlining of byteswap operations" and ef3fb2422f: "scsi: fc: use
get/put_unaligned64 for wwn access") that each try to sidestep distinct
problems with gcc behavior (code growth and increased stack usage).
Unfortunately after both have been applied, a more serious gcc bug has
been uncovered, leading to incorrect object code that discards part of a
function and causes undefined behavior.
As part of this problem is how __builtin_constant_p gets evaluated on an
argument passed by reference into an inline function, this avoids the
use of __builtin_constant_p() for all architectures that set
CONFIG_ARCH_USE_BUILTIN_BSWAP. Most architectures do not set
ARCH_SUPPORTS_OPTIMIZED_INLINING, which means they probably do not
suffer from the problem in the qla2xxx driver, but they might still run
into it elsewhere.
Both of the original workarounds were only merged in the 4.6 kernel, and
the bug that is fixed by this patch should only appear if both are
there, so we probably don't need to backport the fix. On the other
hand, it works by simplifying the code path and should not have any
negative effects.
[arnd@arndb.de: fix older gcc warnings]
(http://lkml.kernel.org/r/12243652.bxSxEgjgfk@wuerfel)
Link: https://lkml.org/lkml/headers/2016/4/12/1103
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66122
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70232
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70646
Fixes: e3bde9568d ("include/linux/unaligned: force inlining of byteswap operations")
Fixes: ef3fb2422f ("scsi: fc: use get/put_unaligned64 for wwn access")
Link: http://lkml.kernel.org/r/1780465.XdtPJpi8Tt@wuerfel
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Tested-by: Josh Poimboeuf <jpoimboe@redhat.com> # on gcc-5.3
Tested-by: Quinn Tran <quinn.tran@qlogic.com>
Cc: Martin Jambor <mjambor@suse.cz>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
Cc: James Bottomley <James.Bottomley@hansenpartnership.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Himanshu Madhani <himanshu.madhani@qlogic.com>
Cc: Jan Hubicka <hubicka@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Recently, we allow to save the stacktrace whose hashed value is 0. It
causes the problem that stackdepot could return 0 even if in success.
User of stackdepot cannot distinguish whether it is success or not so we
need to solve this problem. In this patch, 1 bit are added to handle
and make valid handle none 0 by setting this bit. After that, valid
handle will not be 0 and 0 handle will represent failure correctly.
Fixes: 33334e2576 ("lib/stackdepot.c: allow the stack trace hash to be zero")
Link: http://lkml.kernel.org/r/1462252403-1106-1-git-send-email-iamjoonsoo.kim@lge.com
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Assume memory47 is the last online block left in node1. This will hang:
# echo offline > /sys/devices/system/node/node1/memory47/state
After a couple of minutes, the following pops up in dmesg:
INFO: task bash:957 blocked for more than 120 seconds.
Not tainted 4.6.0-rc6+ #6
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
bash D ffff8800b7adbaf8 0 957 951 0x00000000
Call Trace:
schedule+0x35/0x80
schedule_timeout+0x1ac/0x270
wait_for_completion+0xe1/0x120
kthread_stop+0x4f/0x110
kcompactd_stop+0x26/0x40
__offline_pages.constprop.28+0x7e6/0x840
offline_pages+0x11/0x20
memory_block_action+0x73/0x1d0
memory_subsys_offline+0x47/0x60
device_offline+0x86/0xb0
store_mem_state+0xda/0xf0
dev_attr_store+0x18/0x30
sysfs_kf_write+0x37/0x40
kernfs_fop_write+0x11d/0x170
__vfs_write+0x37/0x120
vfs_write+0xa9/0x1a0
SyS_write+0x55/0xc0
entry_SYSCALL_64_fastpath+0x1a/0xa4
kcompactd is waiting for kcompactd_max_order > 0 when it's woken up to
actually exit. Check kthread_should_stop() to break out of the wait.
Fixes: 698b1b306 ("mm, compaction: introduce kcompactd").
Reported-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Tested-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Since the wildcard at the end of OF module aliases is gone, autoloading
of modules that don't match a device's last (most generic) compatible
value fails.
For example the CODA960 VPU on i.MX6Q has the SoC specific compatible
"fsl,imx6q-vpu" and the generic compatible "cnm,coda960". Since the
driver currently only works with knowledge about the SoC specific
integration, it doesn't list "cnm,cod960" in the module device table.
This results in the device compatible
"of:NvpuT<NULL>Cfsl,imx6q-vpuCcnm,coda960" not matching the module alias
"of:N*T*Cfsl,imx6q-vpu" anymore, whereas before commit 2f632369ab
("modpost: don't add a trailing wildcard for OF module aliases") it
matched the module alias "of:N*T*Cfsl,imx6q-vpu*".
This patch adds two module aliases for each compatible, one without the
wildcard and one with "C*" appended.
$ modinfo coda | grep imx6q
alias: of:N*T*Cfsl,imx6q-vpuC*
alias: of:N*T*Cfsl,imx6q-vpu
Fixes: 2f632369ab ("modpost: don't add a trailing wildcard for OF module aliases")
Link: http://lkml.kernel.org/r/1462203339-15340-1-git-send-email-p.zabel@pengutronix.de
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Cc: Javier Martinez Canillas <javier@osg.samsung.com>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Sjoerd Simons <sjoerd.simons@collabora.co.uk>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: <stable@vger.kernel.org> [4.5+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If /proc/<PID>/environ gets read before the envp[] array is fully set up
in create_{aout,elf,elf_fdpic,flat}_tables(), we might end up trying to
read more bytes than are actually written, as env_start will already be
set but env_end will still be zero, making the range calculation
underflow, allowing to read beyond the end of what has been written.
Fix this as it is done for /proc/<PID>/cmdline by testing env_end for
zero. It is, apparently, intentionally set last in create_*_tables().
This bug was found by the PaX size_overflow plugin that detected the
arithmetic underflow of 'this_len = env_end - (env_start + src)' when
env_end is still zero.
The expected consequence is that userland trying to access
/proc/<PID>/environ of a not yet fully set up process may get
inconsistent data as we're in the middle of copying in the environment
variables.
Fixes: https://forums.grsecurity.net/viewtopic.php?f=3&t=4363
Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=116461
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Emese Revfy <re.emese@gmail.com>
Cc: Pax Team <pageexec@freemail.hu>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Mateusz Guzik <mguzik@redhat.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Jarod Wilson <jarod@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Instead of using "zswap" as the name for all zpools created, add an
atomic counter and use "zswap%x" with the counter number for each zpool
created, to provide a unique name for each new zpool.
As zsmalloc, one of the zpool implementations, requires/expects a unique
name for each pool created, zswap should provide a unique name. The
zsmalloc pool creation does not fail if a new pool with a conflicting
name is created, unless CONFIG_ZSMALLOC_STAT is enabled; in that case,
zsmalloc pool creation fails with -ENOMEM. Then zswap will be unable to
change its compressor parameter if its zpool is zsmalloc; it also will
be unable to change its zpool parameter back to zsmalloc, if it has any
existing old zpool using zsmalloc with page(s) in it. Attempts to
change the parameters will result in failure to create the zpool. This
changes zswap to provide a unique name for each zpool creation.
Fixes: f1c54846ee ("zswap: dynamic pool creation")
Signed-off-by: Dan Streetman <ddstreet@ieee.org>
Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Dan Streetman <dan.streetman@canonical.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After the THP refcounting change, obtaining a compound pages from
get_user_pages() no longer allows us to assume the entire compound page
is immediately mappable from a secondary MMU.
A secondary MMU doesn't want to call get_user_pages() more than once for
each compound page, in order to know if it can map the whole compound
page. So a secondary MMU needs to know from a single get_user_pages()
invocation when it can map immediately the entire compound page to avoid
a flood of unnecessary secondary MMU faults and spurious
atomic_inc()/atomic_dec() (pages don't have to be pinned by MMU notifier
users).
Ideally instead of the page->_mapcount < 1 check, get_user_pages()
should return the granularity of the "page" mapping in the "mm" passed
to get_user_pages(). However it's non trivial change to pass the "pmd"
status belonging to the "mm" walked by get_user_pages up the stack (up
to the caller of get_user_pages). So the fix just checks if there is
not a single pte mapping on the page returned by get_user_pages, and in
turn if the caller can assume that the whole compound page is mapped in
the current "mm" (in a pmd_trans_huge()). In such case the entire
compound page is safe to map into the secondary MMU without additional
get_user_pages() calls on the surrounding tail/head pages. In addition
of being faster, not having to run other get_user_pages() calls also
reduces the memory footprint of the secondary MMU fault in case the pmd
split happened as result of memory pressure.
Without this fix after a MADV_DONTNEED (like invoked by QEMU during
postcopy live migration or balloning) or after generic swapping (with a
failure in split_huge_page() that would only result in pmd splitting and
not a physical page split), KVM would map the whole compound page into
the shadow pagetables, despite regular faults or userfaults (like
UFFDIO_COPY) may map regular pages into the primary MMU as result of the
pte faults, leading to the guest mode and userland mode going out of
sync and not working on the same memory at all times.
Any other secondary MMU notifier manager (KVM is just one of the many
MMU notifier users) will need the same information if it doesn't want to
run a flood of get_user_pages_fast and it can support multiple
granularity in the secondary MMU mappings, so I think it is justified to
be exposed not just to KVM.
The other option would be to move transparent_hugepage_adjust to
mm/huge_memory.c but that currently has all kind of KVM data structures
in it, so it's definitely not a cut-and-paste work, so I couldn't do a
fix as cleaner as this one for 4.6.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: "Li, Liang Z" <liang.z.li@intel.com>
Cc: Amit Shah <amit.shah@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
/proc/sys/vm/stat_refresh warns nr_isolated_anon and nr_isolated_file go
increasingly negative under compaction: which would add delay when
should be none, or no delay when should delay. The bug in compaction
was due to a recent mmotm patch, but much older instance of the bug was
also noticed in isolate_migratepages_range() which is used for CMA and
gigantic hugepage allocations.
The bug is caused by putback_movable_pages() in an error path
decrementing the isolated counters without them being previously
incremented by acct_isolated(). Fix isolate_migratepages_range() by
removing the error-path putback, thus reaching acct_isolated() with
migratepages still isolated, and leaving putback to caller like most
other places do.
Fixes: edc2ca6124 ("mm, compaction: move pageblock checks up from isolate_migratepages_range()")
[vbabka@suse.cz: expanded the changelog]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Khugepaged attempts to raise min_free_kbytes if its set too low.
However, on boot khugepaged sets min_free_kbytes first from
subsys_initcall(), and then the mm 'core' over-rides min_free_kbytes
after from init_per_zone_wmark_min(), via a module_init() call.
Khugepaged used to use a late_initcall() to set min_free_kbytes (such
that it occurred after the core initialization), however this was
removed when the initialization of min_free_kbytes was integrated into
the starting of the khugepaged thread.
The fix here is simply to invoke the core initialization using a
core_initcall() instead of module_init(), such that the previous
initialization ordering is restored. I didn't restore the
late_initcall() since start_stop_khugepaged() already sets
min_free_kbytes via set_recommended_min_free_kbytes().
This was noticed when we had a number of page allocation failures when
moving a workload to a kernel with this new initialization ordering. On
an 8GB system this restores min_free_kbytes back to 67584 from 11365
when CONFIG_TRANSPARENT_HUGEPAGE=y is set and either
CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y or
CONFIG_TRANSPARENT_HUGEPAGE_MADVISE=y.
Fixes: 79553da293 ("thp: cleanup khugepaged startup")
Signed-off-by: Jason Baron <jbaron@akamai.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
zap_pmd_range()'s CONFIG_DEBUG_VM !rwsem_is_locked(&mmap_sem) BUG() will
be invalid with huge pagecache, in whatever way it is implemented:
truncation of a hugely-mapped file to an unhugely-aligned size would
easily hit it.
(Although anon THP could in principle apply khugepaged to private file
mappings, which are not excluded by the MADV_HUGEPAGE restrictions, in
practice there's a vm_ops check which excludes them, so it never hits
this BUG() - there's no interface to "truncate" an anonymous mapping.)
We could complicate the test, to check i_mmap_rwsem also when there's a
vm_file; but my inclination was to make zap_pmd_range() more readable by
simply deleting this check. A search has shown no report of the issue
in the years since commit e0897d75f0 ("mm, thp: print useful
information when mmap_sem is unlocked in zap_pmd_range") expanded it
from VM_BUG_ON() - though I cannot point to what commit I would say then
fixed the issue.
But there are a couple of other patches now floating around, neither yet
in the tree: let's agree to retain the check as a VM_BUG_ON_VMA(), as
Matthew Wilcox has done; but subject to a vma_is_anonymous() check, as
Kirill Shutemov has done. And let's get this in, without waiting for
any particular huge pagecache implementation to reach the tree.
Matthew said "We can reproduce this BUG() in the current Linus tree with
DAX PMDs".
Signed-off-by: Hugh Dickins <hughd@google.com>
Tested-by: Matthew Wilcox <willy@linux.intel.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andres Lagar-Cavilla <andreslc@google.com>
Cc: Yang Shi <yang.shi@linaro.org>
Cc: Ning Qu <quning@gmail.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Andres Lagar-Cavilla <andreslc@google.com>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
