Drop the mem01 test
PATCH 1/1 Drop the mem01 test
The mem01 test often fails on PPC systems mainly because it doesn't take into account page sizes larger than 4KB. Test scenario review revealed that it's not particularly useful because it doesn't have any more specific goal than filling large amount of memory and the hardcoded allocation limits are too low for modern and future systems. The useful part of this test mostly overlaps with coverage by existing OOM tests.
--- runtest/crashme | 3 - runtest/mm | 1 - testcases/kernel/mem/.gitignore | 1 - testcases/kernel/mem/mem/mem01.c | 249 ------------------------------- 4 files changed, 254 deletions(-) delete mode 100644 testcases/kernel/mem/mem/mem01.c
runtest/crashme +0 -3
@@ -10,9 +10,6 @@ crash01 crash01 crash02 crash02 -v 2 # Generate random syscalls and execute them, less probability # to hose your system, but still. -mem01 mem01 -r -# Memory eater. Loves to be run in parallel with other programs. -# May panic on buggy systems if the OOM killer was not fast enough :-) fork12 fork12 # Fork as many children as possible. On systems with lots of memory # and kernels prior to 2.4.19, this can hang the system by using up all pids
runtest/mm +0 -1
@@ -20,7 +20,6 @@ mtest06_3 mmap3 -x 0.002 -p # Remains diabled till the infinite loop problem is solved #mtest-6_4 shmat1 -x 0.00005 -mem01 mem01 mem02 mem02 mem03 mem03
testcases/kernel/mem/.gitignore +0 -1
@@ -23,7 +23,6 @@ /ksm/ksm04 /ksm/ksm05 /ksm/ksm06 -/mem/mem01 /mem/mem02 /mmapstress/mmap-corruption01 /mmapstress/mmapstress01
testcases/kernel/mem/mem/mem01.c deleted +0 -249
249 lines not on this page · raw message
Signature · 1 line
2.27.0
PATCH 1/1 Drop the mem01 test
Quoted 8 lines · Martin Doucha
The mem01 test often fails on PPC systems mainly because it doesn't take into
account page sizes larger than 4KB. Test scenario review revealed that it's not particularly useful because it doesn't have any more specific goal than filling large amount of memory and the hardcoded allocation limits are too low for modern and future systems. The useful part of this test mostly overlaps with coverage by existing OOM tests.
Agree, we also use some other memory eater or stress tools
do the same testing, this mem01 test way looks not very necessary and older.
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20200807/07e59c58/attachment-0001.htm>
PATCH 1/1 Drop the mem01 test
In reply to Martin Doucha----- Original Message -----
Quoted 8 lines
The mem01 test often fails on PPC systems mainly because it doesn't take into account page sizes larger than 4KB. Test scenario review revealed that it's not particularly useful because it doesn't have any more specific goal than filling large amount of memory and the hardcoded allocation limits are too low for modern and future systems. The useful part of this test mostly overlaps with coverage by existing OOM tests.
Signed-off-by Martin Doucha <[email protected]>
I'm fine with dropping it. Just curious, how is it failing? Do you see it hit TFAIL, OOM or something else?
PATCH 1/1 Drop the mem01 test
Quoted 2 lines · Jan Stancek
I'm fine with dropping it. Just curious, how is it failing? Do you see it hit TFAIL, OOM or something else?
It's getting killed by OOM. The test leaves a hard-coded safety margin of 16MB unallocated. That's 4096 pages on x86_64 but only 256 pages on PPC. Making the safety margin pagesize-dependent would be the obvious fix if we wanted to keep the test.
Signature · 7 lines
Martin Doucha [email protected] QA Engineer for Software Maintenance SUSE LINUX, s.r.o. CORSO IIa Krizikova 148/34 186 00 Prague 8 Czech Republic
PATCH 1/1 Drop the mem01 test
In reply to Jan StancekQuoted 8 lines · Jan Stancek
----- Original Message -----
Quoted 1 line
The mem01 test often fails on PPC systems mainly because it doesn't take
into
Quoted 1 line
account page sizes larger than 4KB. Test scenario review revealed that
it's
Quoted 1 line
not particularly useful because it doesn't have any more specific goal
than
Quoted 8 lines
Quoted 1 line
filling large amount of memory and the hardcoded allocation limits are
too
Quoted 4 lines
low for modern and future systems. The useful part of this test mostly overlaps with coverage by existing OOM tests.
Signed-off-by Martin Doucha <[email protected]>Acked-by Jan Stancek <[email protected]>
Pushed, thanks!
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20200817/d8f0d7f6/attachment.htm>
PATCH 1/1 Drop the mem01 test
Hi Li
Quoted 22 lines
On Mon, Aug 10, 2020 at 4:24 PM Jan Stancek <[email protected] <mailto:[email protected]>> wrote:
----- Original Message -----
> The mem01 test often fails on PPC systems mainly because it
doesn't take into
> account page sizes larger than 4KB. Test scenario review revealed
that it's
> not particularly useful because it doesn't have any more specific
goal than
> filling large amount of memory and the hardcoded allocation
limits are too
> low for modern and future systems. The useful part of this test
mostly
> overlaps with coverage by existing OOM tests.
>
> Signed-off-by: Martin Doucha <[email protected]
<mailto:[email protected]>>
Quoted 4 lines
Acked-by Jan Stancek <[email protected] <mailto:[email protected]>>Pushed, thanks!
[root@localhost ltp]# git grep mem01 testcases/kernel/syscalls/ioctl/ioctl_sg01.c: * a read-accessible SCSI-compatible device (e.g. SATA disk). Running mem01 [root@localhost ltp]#
ioctl_sg01.c has such description " Running mem01
* test program before this one may increase the chance of successfully
* reproducing the bug."
I guess we should also remove this or use oom testcases can also increase this rate(I don't try ).
Best Regards
Yang Xu
Quoted 6 lines
Signature · 3 lines
-- Regards, Li Wang
PATCH 1/1 Drop the mem01 test
Quoted 35 lines · Yang Xu
Hi Li
Quoted 22 lines
On Mon, Aug 10, 2020 at 4:24 PM Jan Stancek <[email protected] <mailto:[email protected]>> wrote:
----- Original Message -----
> The mem01 test often fails on PPC systems mainly because it
doesn't take into
> account page sizes larger than 4KB. Test scenario review revealed
that it's
> not particularly useful because it doesn't have any more specific
goal than
> filling large amount of memory and the hardcoded allocation
limits are too
> low for modern and future systems. The useful part of this test
mostly
> overlaps with coverage by existing OOM tests.
>
> Signed-off-by: Martin Doucha <[email protected]
<mailto:[email protected]>>Quoted 1 line
Quoted 3 lines
Pushed, thanks!
[root@localhost ltp]# git grep mem01
testcases/kernel/syscalls/ioctl/ioctl_sg01.c: * a read-accessible SCSI-compatible device (e.g. SATA disk). Running mem01 [root@localhost ltp]#
Quoted 7 lines
ioctl_sg01.chas such description " Running mem01
* test program before this one may increase the chance of successfully
* reproducing the bug."I guess we should also remove this or use oom testcases can also increase this rate(I don't try ).
Good catch, I will help submit a patch to fix that code comment.
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20200817/5e323fd6/attachment.htm>
PATCH 1/1 Drop the mem01 test
In reply to Yang XuQuoted 11 lines · Yang Xu
[root@localhost ltp]# git grep mem01
testcases/kernel/syscalls/ioctl/ioctl_sg01.c: * a read-accessible SCSI-compatible device (e.g. SATA disk). Running mem01 [root@localhost ltp]#
ioctl_sg01.chas such description " Running mem01
?* test program before this one may increase the chance of successfully ?* reproducing the bug."I guess we should also remove this or use oom testcases can also increase this rate(I don't try ).
OOM testcases will work fine. ioctl_sg01 is a CVE test checking for kernel information disclosure bug so it just needs the kernel to reuse some dirty block of memory for the device buffer that then gets copied to userspace.
Maybe I should improve the test by dirtying some kernel memory myself in the test setup...
Signature · 7 lines
Martin Doucha [email protected] QA Engineer for Software Maintenance SUSE LINUX, s.r.o. CORSO IIa Krizikova 148/34 186 00 Prague 8 Czech Republic