2022-10-24 19:22:49

by Gautam Menghani

[permalink] [raw]
Subject: [PATCH v2] mm/khugepaged: add tracepoint to collapse_file()

In the file mm/khugepaged.c, a TODO in the function collapse_file() asks
to add tracepoints. Add the tracepoint named "mm_khugepaged_collapse_file".

Signed-off-by: Gautam Menghani <[email protected]>
---
changes in v2:
1. Move filename dereference out of calling path (Steven Rostedt).
2. Capture is_shmem and index (Zach O'Keefe)

include/trace/events/huge_memory.h | 38 ++++++++++++++++++++++++++++++
mm/khugepaged.c | 4 +++-
2 files changed, 41 insertions(+), 1 deletion(-)

diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
index 935af4947917..472238d0e3a3 100644
--- a/include/trace/events/huge_memory.h
+++ b/include/trace/events/huge_memory.h
@@ -203,5 +203,43 @@ TRACE_EVENT(mm_khugepaged_scan_file,
__print_symbolic(__entry->result, SCAN_STATUS))
);

+TRACE_EVENT(mm_khugepaged_collapse_file,
+ TP_PROTO(struct mm_struct *mm, struct page *hpage, pgoff_t index,
+ bool is_shmem, unsigned long addr, struct file *file,
+ int nr, int result),
+ TP_ARGS(mm, hpage, index, is_shmem, addr, file, nr, result),
+ TP_STRUCT__entry(
+ __field(struct mm_struct *, mm)
+ __field(unsigned long, hpfn)
+ __field(pgoff_t, index)
+ __field(bool, is_shmem)
+ __field(unsigned long, addr)
+ __string(filename, file->f_path.dentry->d_iname)
+ __field(int, nr)
+ __field(int, result)
+ ),
+
+ TP_fast_assign(
+ __entry->mm = mm;
+ __entry->hpfn = hpage ? page_to_pfn(hpage) : -1;
+ __entry->index = index;
+ __entry->is_shmem = is_shmem;
+ __entry->addr = addr;
+ __assign_str(filename, file->f_path.dentry->d_iname);
+ __entry->nr = nr;
+ __entry->result = result;
+ ),
+
+ TP_printk("mm=%p, hpage_pfn=0x%lx, index=%ld, is_shmem=%d, addr=%ld, filename=%s, nr=%d, result=%s",
+ __entry->mm,
+ __entry->hpfn,
+ __entry->index,
+ __entry->is_shmem,
+ __entry->addr,
+ __get_str(filename),
+ __entry->nr,
+ __print_symbolic(__entry->result, SCAN_STATUS))
+);
+
#endif /* __HUGE_MEMORY_H */
#include <trace/define_trace.h>
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 4734315f7940..977c0ff82c46 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -2059,7 +2059,9 @@ static int collapse_file(struct mm_struct *mm, unsigned long addr,
mem_cgroup_uncharge(page_folio(hpage));
put_page(hpage);
}
- /* TODO: tracepoints */
+
+ trace_mm_khugepaged_collapse_file(mm, hpage, index, is_shmem,
+ addr, file, nr, result);
return result;
}

--
2.34.1


2022-10-24 23:33:54

by Andrew Morton

[permalink] [raw]
Subject: Re: [PATCH v2] mm/khugepaged: add tracepoint to collapse_file()

On Mon, 24 Oct 2022 23:05:58 +0530 Gautam Menghani <[email protected]> wrote:

> In the file mm/khugepaged.c, a TODO in the function collapse_file() asks
> to add tracepoints. Add the tracepoint named "mm_khugepaged_collapse_file".

This isn't a very satisfying explanation for changing the kernel. Maybe
the comment is stale are this tracepoint is unneeded.

Please explain afresh how this addition benefits kernel users?

2022-10-25 05:26:56

by Gautam Menghani

[permalink] [raw]
Subject: Re: [PATCH v2] mm/khugepaged: add tracepoint to collapse_file()

On Mon, Oct 24, 2022 at 01:17:06PM -0700, Andrew Morton wrote:
> On Mon, 24 Oct 2022 23:05:58 +0530 Gautam Menghani <[email protected]> wrote:
>
> > In the file mm/khugepaged.c, a TODO in the function collapse_file() asks
> > to add tracepoints. Add the tracepoint named "mm_khugepaged_collapse_file".
>
> This isn't a very satisfying explanation for changing the kernel. Maybe
> the comment is stale are this tracepoint is unneeded.
>
> Please explain afresh how this addition benefits kernel users?
>
The function collapse_file() is called by the function hpage_collapse_scan_file().
Without a tracepoint in collapse_file(), we won't know if it was called or not and as a result,
we also won't know if it returned successfully or not. Also, as Zach mentioned earlier [1]:

there are a few scan result codes that overlap between hpage_collapse_scan_file() and those
possibly returned in collapse_file() such that, if we only have the one tracepoint in
hpage_collapse_scan_file(), it could be ambiguous what callsite the error path stemmed from.

[1]:https://lore.kernel.org/lkml/CAAa6QmSKtj6T2dW1tkg5_HVj2+rXj5inOLdEzr0MkJzQxxcPXQ@mail.gmail.com/

Please do let me know if a v3 is needed.

Thanks,
Gautam

2022-10-25 19:15:18

by Zach O'Keefe

[permalink] [raw]
Subject: Re: [PATCH v2] mm/khugepaged: add tracepoint to collapse_file()

On Oct 25 10:35, Gautam Menghani wrote:
> On Mon, Oct 24, 2022 at 01:17:06PM -0700, Andrew Morton wrote:
> > On Mon, 24 Oct 2022 23:05:58 +0530 Gautam Menghani <[email protected]> wrote:
> >
> > > In the file mm/khugepaged.c, a TODO in the function collapse_file() asks
> > > to add tracepoints. Add the tracepoint named "mm_khugepaged_collapse_file".
> >
> > This isn't a very satisfying explanation for changing the kernel. Maybe
> > the comment is stale are this tracepoint is unneeded.
> >
> > Please explain afresh how this addition benefits kernel users?
> >
> The function collapse_file() is called by the function hpage_collapse_scan_file().
> Without a tracepoint in collapse_file(), we won't know if it was called or not and as a result,
> we also won't know if it returned successfully or not. Also, as Zach mentioned earlier [1]:
>
> there are a few scan result codes that overlap between hpage_collapse_scan_file() and those
> possibly returned in collapse_file() such that, if we only have the one tracepoint in
> hpage_collapse_scan_file(), it could be ambiguous what callsite the error path stemmed from.
>
> [1]:https://lore.kernel.org/lkml/CAAa6QmSKtj6T2dW1tkg5_HVj2+rXj5inOLdEzr0MkJzQxxcPXQ@mail.gmail.com/
>
> Please do let me know if a v3 is needed.
>
> Thanks,
> Gautam

Thanks Guatam,

The ambiguous codes in particular are:

SCAN_PTE_MAPPED_HUGEPAGE
SCAN_PAGE_COMPOUND
SCAN_PAGE_COUNT

; properties of a page that are checked before/after the page is
locked/isolated. This personally hasn't been an issue for me, as someone who's
spent considerable time in these codepaths over the last little while. There
are certainly other codes (like SCAN_FAIL) which are ambiguous within functions
(and likewise haven't been an issue) so this reason alone isn't particularly
motivating.

However, some of the extra information (is_shmem, index), at times, would have
been useful if was already available -- but it was never enough of a pain to
force my hand to put this tracepoint in.

If the tracepoint is staying, then 2 nits:

CHECK: Alignment should match open parenthesis
#74: FILE: mm/khugepaged.c:2064:
+ trace_mm_khugepaged_collapse_file(mm, hpage, index, is_shmem,
+ addr, file, nr, result);

and, apply Steven's advice to trace_mm_khugepaged_scan_file() for consistency.

Thanks,
Zach

2022-10-26 05:10:23

by Gautam Menghani

[permalink] [raw]
Subject: Re: [PATCH v2] mm/khugepaged: add tracepoint to collapse_file()

On Tue, Oct 25, 2022 at 11:20:40AM -0700, Zach O'Keefe wrote:
> On Oct 25 10:35, Gautam Menghani wrote:
> > On Mon, Oct 24, 2022 at 01:17:06PM -0700, Andrew Morton wrote:
> > > On Mon, 24 Oct 2022 23:05:58 +0530 Gautam Menghani <[email protected]> wrote:
> > >
> > > > In the file mm/khugepaged.c, a TODO in the function collapse_file() asks
> > > > to add tracepoints. Add the tracepoint named "mm_khugepaged_collapse_file".
> > >
> > > This isn't a very satisfying explanation for changing the kernel. Maybe
> > > the comment is stale are this tracepoint is unneeded.
> > >
> > > Please explain afresh how this addition benefits kernel users?
> > >
> > The function collapse_file() is called by the function hpage_collapse_scan_file().
> > Without a tracepoint in collapse_file(), we won't know if it was called or not and as a result,
> > we also won't know if it returned successfully or not. Also, as Zach mentioned earlier [1]:
> >
> > there are a few scan result codes that overlap between hpage_collapse_scan_file() and those
> > possibly returned in collapse_file() such that, if we only have the one tracepoint in
> > hpage_collapse_scan_file(), it could be ambiguous what callsite the error path stemmed from.
> >
> > [1]:https://lore.kernel.org/lkml/CAAa6QmSKtj6T2dW1tkg5_HVj2+rXj5inOLdEzr0MkJzQxxcPXQ@mail.gmail.com/
> >
> > Please do let me know if a v3 is needed.
> >
> > Thanks,
> > Gautam
>
> Thanks Guatam,
>
> The ambiguous codes in particular are:
>
> SCAN_PTE_MAPPED_HUGEPAGE
> SCAN_PAGE_COMPOUND
> SCAN_PAGE_COUNT
>
> ; properties of a page that are checked before/after the page is
> locked/isolated. This personally hasn't been an issue for me, as someone who's
> spent considerable time in these codepaths over the last little while. There
> are certainly other codes (like SCAN_FAIL) which are ambiguous within functions
> (and likewise haven't been an issue) so this reason alone isn't particularly
> motivating.
>
> However, some of the extra information (is_shmem, index), at times, would have
> been useful if was already available -- but it was never enough of a pain to
> force my hand to put this tracepoint in.
>
> If the tracepoint is staying, then 2 nits:
>
> CHECK: Alignment should match open parenthesis
> #74: FILE: mm/khugepaged.c:2064:
> + trace_mm_khugepaged_collapse_file(mm, hpage, index, is_shmem,
> + addr, file, nr, result);
>
> and, apply Steven's advice to trace_mm_khugepaged_scan_file() for consistency.
Yes I've sent a patch to refactor the mm_khugepaged_scan_file() tracepoint -
lore.kernel.org/linux-mm/[email protected]/T/#u