2021-08-25 14:04:44

by George G. Davis

Subject: [RFC][PATCH] selftests/vm/transhuge-stress: fix ram size thinko

From: "George G. Davis" <[email protected]>

When executing transhuge-stress with an argument to specify the virtual
memory size for testing, the ram size is reported as 0, e.g.

transhuge-stress 384
thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 0 MiB of ram
thp-mmap: 0.184 s/loop, 0.957 ms/page, 2090.265 MiB/s 192 succeed, 0 failed

This appears to be due to a thinko in commit 0085d61fe05e
("selftests/vm/transhuge-stress: stress test for memory compaction"),
where, at a guess, the intent was to base "xyz MiB of ram" on `ram`
size. Here are results after using `ram` size:

thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 14 MiB of ram

Fixes: 0085d61fe05e ("selftests/vm/transhuge-stress: stress test for memory compaction")
Signed-off-by: George G. Davis <[email protected]>
---
tools/testing/selftests/vm/transhuge-stress.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/selftests/vm/transhuge-stress.c b/tools/testing/selftests/vm/transhuge-stress.c
index fd7f1b4a96f9..5e4c036f6ad3 100644
--- a/tools/testing/selftests/vm/transhuge-stress.c
+++ b/tools/testing/selftests/vm/transhuge-stress.c
@@ -79,7 +79,7 @@ int main(int argc, char **argv)

warnx("allocate %zd transhuge pages, using %zd MiB virtual memory"
" and %zd MiB of ram", len >> HPAGE_SHIFT, len >> 20,
- len >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));
+ ram >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));

pagemap_fd = open("/proc/self/pagemap", O_RDONLY);
if (pagemap_fd < 0)
--
2.17.1

2021-09-28 18:08:18

by George G. Davis

Subject: Re: [RFC][PATCH] selftests/vm/transhuge-stress: fix ram size thinko

On Wed, Aug 25, 2021 at 09:58:43AM -0400, George G. Davis wrote:
> From: "George G. Davis" <[email protected]>
>
> When executing transhuge-stress with an argument to specify the virtual
> memory size for testing, the ram size is reported as 0, e.g.
>
> transhuge-stress 384
> thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 0 MiB of ram
> thp-mmap: 0.184 s/loop, 0.957 ms/page, 2090.265 MiB/s 192 succeed, 0 failed
>
> This appears to be due to a thinko in commit 0085d61fe05e
> ("selftests/vm/transhuge-stress: stress test for memory compaction"),
> where, at a guess, the intent was to base "xyz MiB of ram" on `ram`
> size. Here are results after using `ram` size:
>
> thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 14 MiB of ram
>
> Fixes: 0085d61fe05e ("selftests/vm/transhuge-stress: stress test for memory compaction")
> Signed-off-by: George G. Davis <[email protected]>
> ---
> tools/testing/selftests/vm/transhuge-stress.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/vm/transhuge-stress.c b/tools/testing/selftests/vm/transhuge-stress.c
> index fd7f1b4a96f9..5e4c036f6ad3 100644
> --- a/tools/testing/selftests/vm/transhuge-stress.c
> +++ b/tools/testing/selftests/vm/transhuge-stress.c
> @@ -79,7 +79,7 @@ int main(int argc, char **argv)
>
> warnx("allocate %zd transhuge pages, using %zd MiB virtual memory"
> " and %zd MiB of ram", len >> HPAGE_SHIFT, len >> 20,
> - len >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));
> + ram >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));

Ping, any comments on this?

For the record, as noted, the value reported for "ram size" appears to
use the wrong variable, `len`, where `ram` should be used instead.

>
> pagemap_fd = open("/proc/self/pagemap", O_RDONLY);
> if (pagemap_fd < 0)
> --
> 2.17.1
>

--
Regards,
George

2021-10-16 22:17:58

by Shuah Khan

Subject: Re: [RFC][PATCH] selftests/vm/transhuge-stress: fix ram size thinko

On 8/25/21 7:58 AM, George G. Davis wrote:
> From: "George G. Davis" <[email protected]>
>
> When executing transhuge-stress with an argument to specify the virtual
> memory size for testing, the ram size is reported as 0, e.g.
>
> transhuge-stress 384
> thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 0 MiB of ram
> thp-mmap: 0.184 s/loop, 0.957 ms/page, 2090.265 MiB/s 192 succeed, 0 failed
>
> This appears to be due to a thinko in commit 0085d61fe05e
> ("selftests/vm/transhuge-stress: stress test for memory compaction"),
> where, at a guess, the intent was to base "xyz MiB of ram" on `ram`
> size. Here are results after using `ram` size:
>
> thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 14 MiB of ram
>
> Fixes: 0085d61fe05e ("selftests/vm/transhuge-stress: stress test for memory compaction")
> Signed-off-by: George G. Davis <[email protected]>
> ---
> tools/testing/selftests/vm/transhuge-stress.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/vm/transhuge-stress.c b/tools/testing/selftests/vm/transhuge-stress.c
> index fd7f1b4a96f9..5e4c036f6ad3 100644
> --- a/tools/testing/selftests/vm/transhuge-stress.c
> +++ b/tools/testing/selftests/vm/transhuge-stress.c
> @@ -79,7 +79,7 @@ int main(int argc, char **argv)
>
> warnx("allocate %zd transhuge pages, using %zd MiB virtual memory"
> " and %zd MiB of ram", len >> HPAGE_SHIFT, len >> 20,
> - len >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));
> + ram >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));
>
> pagemap_fd = open("/proc/self/pagemap", O_RDONLY);
> if (pagemap_fd < 0)
>

Sorry for the delay on this. The change looks good to me.

Andrew! Would you like me to take this through kselftest tree?

thanks,
-- Shuah

2021-10-17 22:04:18

by Andrew Morton

Subject: Re: [RFC][PATCH] selftests/vm/transhuge-stress: fix ram size thinko

On Fri, 15 Oct 2021 09:38:24 -0600 Shuah Khan <[email protected]> wrote:

> On 8/25/21 7:58 AM, George G. Davis wrote:
> > From: "George G. Davis" <[email protected]>
> >
> > When executing transhuge-stress with an argument to specify the virtual
> > memory size for testing, the ram size is reported as 0, e.g.
> >
> > transhuge-stress 384
> > thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 0 MiB of ram
> > thp-mmap: 0.184 s/loop, 0.957 ms/page, 2090.265 MiB/s 192 succeed, 0 failed
> >
> > This appears to be due to a thinko in commit 0085d61fe05e
> > ("selftests/vm/transhuge-stress: stress test for memory compaction"),
> > where, at a guess, the intent was to base "xyz MiB of ram" on `ram`
> > size. Here are results after using `ram` size:
> >
> > thp-mmap: allocate 192 transhuge pages, using 384 MiB virtual memory and 14 MiB of ram
> >
> > Fixes: 0085d61fe05e ("selftests/vm/transhuge-stress: stress test for memory compaction")
> > Signed-off-by: George G. Davis <[email protected]>
> > ---
> > tools/testing/selftests/vm/transhuge-stress.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/testing/selftests/vm/transhuge-stress.c b/tools/testing/selftests/vm/transhuge-stress.c
> > index fd7f1b4a96f9..5e4c036f6ad3 100644
> > --- a/tools/testing/selftests/vm/transhuge-stress.c
> > +++ b/tools/testing/selftests/vm/transhuge-stress.c
> > @@ -79,7 +79,7 @@ int main(int argc, char **argv)
> >
> > warnx("allocate %zd transhuge pages, using %zd MiB virtual memory"
> > " and %zd MiB of ram", len >> HPAGE_SHIFT, len >> 20,
> > - len >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));
> > + ram >> (20 + HPAGE_SHIFT - PAGE_SHIFT - 1));
> >
> > pagemap_fd = open("/proc/self/pagemap", O_RDONLY);
> > if (pagemap_fd < 0)
> >
>
> Sorry for the delay on this. The change looks good to me.

Konstantin, coould you please take a look?

> Andrew! Would you like me to take this through kselftest tree?

Is OK thanks - I'll add it to my mm/thp pile.