[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <608379FD-DCBA-414A-A0A1-58E0CAECBDEB@nvidia.com>
Date: Thu, 14 Aug 2025 11:47:29 -0400
From: Zi Yan <ziy@...dia.com>
To: Usama Arif <usamaarif642@...il.com>
Cc: Andrew Morton <akpm@...ux-foundation.org>, david@...hat.com,
linux-mm@...ck.org, linux-fsdevel@...r.kernel.org, corbet@....net,
rppt@...nel.org, surenb@...gle.com, mhocko@...e.com, hannes@...xchg.org,
baohua@...nel.org, shakeel.butt@...ux.dev, riel@...riel.com,
laoar.shao@...il.com, dev.jain@....com, baolin.wang@...ux.alibaba.com,
npache@...hat.com, lorenzo.stoakes@...cle.com, Liam.Howlett@...cle.com,
ryan.roberts@....com, vbabka@...e.cz, jannh@...gle.com,
Arnd Bergmann <arnd@...db.de>, sj@...nel.org, linux-kernel@...r.kernel.org,
linux-doc@...r.kernel.org, kernel-team@...a.com
Subject: Re: [PATCH v4 4/7] docs: transhuge: document process level THP
controls
On 13 Aug 2025, at 9:55, Usama Arif wrote:
> This includes the PR_SET_THP_DISABLE/PR_GET_THP_DISABLE pair of
> prctl calls as well the newly introduced PR_THP_DISABLE_EXCEPT_ADVISED
> flag for the PR_SET_THP_DISABLE prctl call.
>
> Signed-off-by: Usama Arif <usamaarif642@...il.com>
> ---
> Documentation/admin-guide/mm/transhuge.rst | 37 ++++++++++++++++++++++
> 1 file changed, 37 insertions(+)
>
> diff --git a/Documentation/admin-guide/mm/transhuge.rst b/Documentation/admin-guide/mm/transhuge.rst
> index 370fba1134606..fa8242766e430 100644
> --- a/Documentation/admin-guide/mm/transhuge.rst
> +++ b/Documentation/admin-guide/mm/transhuge.rst
> @@ -225,6 +225,43 @@ to "always" or "madvise"), and it'll be automatically shutdown when
> PMD-sized THP is disabled (when both the per-size anon control and the
> top-level control are "never")
>
> +process THP controls
> +--------------------
> +
> +A process can control its own THP behaviour using the ``PR_SET_THP_DISABLE``
> +and ``PR_GET_THP_DISABLE`` pair of prctl(2) calls. The THP behaviour set using
> +``PR_SET_THP_DISABLE`` is inherited across fork(2) and execve(2). These calls
> +support the following arguments::
> +
> + prctl(PR_SET_THP_DISABLE, 1, 0, 0, 0):
> + This will disable THPs completely for the process, irrespective
> + of global THP controls or MADV_COLLAPSE.
> +
> + prctl(PR_SET_THP_DISABLE, 1, PR_THP_DISABLE_EXCEPT_ADVISED, 0, 0):
> + This will disable THPs for the process except when the usage of THPs is
> + advised. Consequently, THPs will only be used when:
> + - Global THP controls are set to "always" or "madvise" and
> + the area either has VM_HUGEPAGE set (e.g., due do MADV_HUGEPAGE) or
> + MADV_COLLAPSE is used.
It is better to change the above sentence to:
madvise(..., MADV_HUGEPAGE) or madvise(..., MADV_COLLAPSE) is used.
Since this document is for sysadmin, who does not need to know the implementation
details like VM_HUGEPAGE. And I do not see any kernel internal is mentioned
in the rest of the document.
> + - Global THP controls are set to "never" and MADV_COLLAPSE is used. This
> + is the same behavior as if THPs would not be disabled on a process
> + level.
> + Note that MADV_COLLAPSE is currently always rejected if VM_NOHUGEPAGE is
> + set on an area.
The same for the above sentence.
Something like:
Note that MADV_COLLAPSE is always rejected if madvise(..., MADV_NOHUGEPAGE) is
used.
> +
> + prctl(PR_SET_THP_DISABLE, 0, 0, 0, 0):
> + This will re-enabled THPs for the process, as if they would never have
s/re-enabled/re-enable/
> + been disabled. Whether THPs will actually be used depends on global THP
> + controls.
and madvise() calls.
> +
> + prctl(PR_GET_THP_DISABLE, 0, 0, 0, 0):
> + This returns a value whose bit indicate how THP-disable is configured:
s/bit/bits
> + Bits
> + 1 0 Value Description
> + |0|0| 0 No THP-disable behaviour specified.
> + |0|1| 1 THP is entirely disabled for this process.
> + |1|1| 3 THP-except-advised mode is set for this process.
> +
> Khugepaged controls
> -------------------
>
> --
> 2.47.3
Otherwise, LGTM. Reviewed-by: Zi Yan <ziy@...dia.com>
Best Regards,
Yan, Zi
Powered by blists - more mailing lists