[LTP] [PATCH] shmget02: reduce the shmmax test value in compat mode

[Li Wang]

On Tue, May 30, 2023 at 8:46 PM Martin Doucha <mdoucha@suse.cz> wrote:

> Hi,
>
> On 30. 05. 23 11:30, Li Wang wrote:
> > As Arnd Bergmann pointed out that SHMMAX being defined as
> > (ULONG_MAX - (1UL << 24)), so the kernel would likely use
> > a large 64-bit value, while the 32-bit user space uses a
> > much smaller limit.
> >
> > It finally results in ENOMEM failure:
> >    shmget02.c:95: TFAIL: shmget(1644199826, 4278190080, 1536)
> >                   expected EINVAL: ENOMEM (12)
> >
> > With suggest by Manfred Spraul we could reduce the value
> > of '/proc/sys/kernel/shmmax' in compat mode and only test
> > the overflow behavior with default+1.
> >
> > Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
> > Signed-off-by: Li Wang <liwang@redhat.com>
> > Cc: Naresh Kamboju <naresh.kamboju@linaro.org>
> > Cc: Arnd Bergmann <arnd@arndb.de>
> > Cc: Manfred Spraul <manfred@colorfullife.com>
> > ---
> >   testcases/kernel/syscalls/ipc/shmget/shmget02.c | 10 ++++++++++
> >   1 file changed, 10 insertions(+)
> >
> > diff --git a/testcases/kernel/syscalls/ipc/shmget/shmget02.c
> b/testcases/kernel/syscalls/ipc/shmget/shmget02.c
> > index 7989ef33e..faf633ad4 100644
> > --- a/testcases/kernel/syscalls/ipc/shmget/shmget02.c
> > +++ b/testcases/kernel/syscalls/ipc/shmget/shmget02.c
> > @@ -56,7 +56,11 @@ static struct tcase {
> >       {&shmkey1, SHM_SIZE, IPC_EXCL, 0, 0, ENOENT},
> >       {&shmkey, SHM_SIZE, IPC_CREAT | IPC_EXCL, 0, 0, EEXIST},
> >       {&shmkey1, SHMMIN - 1, IPC_CREAT | IPC_EXCL, 0, 0, EINVAL},
> > +#ifdef TST_ABI32
> > +     {&shmkey1, 8192 + 1, IPC_CREAT | IPC_EXCL, 0, 0, EINVAL},
> > +#else
> >       {&shmkey1, SHMMAX + 1, IPC_CREAT | IPC_EXCL, 0, 0, EINVAL},
> > +#endif
> >       {&shmkey, SHM_SIZE * 2, IPC_EXCL, 0, 0, EINVAL},
> >       {&shmkey, SHM_SIZE, SHM_RD, 1, 0, EACCES},
> >       {&shmkey1, SHM_SIZE, IPC_CREAT | SHM_HUGETLB, 0, 1, EPERM},
> > @@ -149,4 +153,10 @@ static struct tst_test test = {
> >       .test = do_test,
> >       .tcnt = ARRAY_SIZE(tcases),
> >       .hugepages = {TST_NO_HUGEPAGES},
> > +#ifdef TST_ABI32
> > +     .save_restore = (const struct tst_path_val[]) {
> > +             {"/proc/sys/kernel/shmmax", "8192", TST_SR_TBROK},
> > +             {}
> > +     },
> > +#endif
> >   };
>
> Could this be done without the conditional compilation? For example
> change the testcase to custom shmmax limit unconditionally or read the
> actual kernel limit from /proc/sys/kernel/shmmax in setup().
>

That could work as well, but I thought that might a bit complicated
to reassign a new value to the two-level pointer 'tc->size'.


-- 
Regards,
Li Wang