From: Nilay Shroff <nilay@linux.ibm.com>
To: Marco Elver <elver@google.com>, Christoph Hellwig <hch@lst.de>
Cc: linux-nvme@lists.infradead.org, linux-kernel@vger.kernel.org,
kbusch@kernel.org, sagi@grimberg.me, axboe@fb.com,
bvanassche@acm.org, gjoyce@linux.ibm.com
Subject: Re: [PATCHv2 07/17] nvme: add Clang context annotations for nvme_subsystem::lock
Date: Tue, 30 Jun 2026 15:23:54 +0530 [thread overview]
Message-ID: <3d7fd5f4-26b1-42a6-89c5-961ca53c51b1@linux.ibm.com> (raw)
In-Reply-To: <CANpmjNMCg_Qo9MVz3n_-gu6wRZp+y2uuuzjAdk+a1b-1J4LeNA@mail.gmail.com>
On 6/30/26 4:42 AM, Marco Elver wrote:
> On Mon, 29 Jun 2026 at 14:47, Christoph Hellwig <hch@lst.de> wrote:
>>
>> On Fri, Jun 26, 2026 at 08:09:52PM +0530, Nilay Shroff wrote:
>>>>
>>> yes correct, but this one is tricky as the list has to be inited
>>> with non-zero value. Another way to fix this is by adding
>>> context_unsafe wrapper around INIT_LIST_HEAD(). Again overuse of
>>> those unsafe wrappers hinders readability.
>>>
>>> Would it make sense to introduce a helper, e.g. INIT_LIST_HEAD_UNSAFE(),
>>> that simply wraps INIT_LIST_HEAD() with context_unsafe()? It would document
>>> that this should only be used in cases where the caller knows the object
>>> has not yet been published and therefore no concurrent access is possible,
>>> such as during object initialization.
>>
>> I'll leave that to Marco and other core people, but I think this
>> would be a lot better than the current version.
>
> Initialization of guarded objects has gone through a few iterations,
> so I don't want to open that can of worms again. I think the
> infrastructure we have now provides various options (the scoped guard
> machinery isn't the only way). You could just write:
>
> /* Initializes unpublished lock-guarded variables. */
> context_unsafe(
> INIT_LIST_HEAD(&subsys->nsheads);
> // ... other guarded var init in same block ...
> );
>
> context_unsafe() has a statement expression inside, so you can have
> multiple statements in one context_unsafe() block.
> And leave the mutex_init() where it was if you don't like the scoped_guard().
>
> Although here you just have 1 INIT_LIST_HEAD() that needs that
> treatment, so "context_unsafe(INIT_LIST_HEAD(...))" would do -- it's
> just 9 chars more vs INIT_LIST_HEAD_UNSAFE(). And as soon as you have
> several guarded variable initializations in a block, you can group
> them as above.
>
> If you introduce INIT_LIST_HEAD_UNSAFE(), that'll create more ways to
> initialize guarded variables. I'm not against it if none of the other
> ways work for you, but expanding the API surface also incurs a
> complexity cost.
>
> The simplest option is just adding __context_unsafe(init) as an
> attribute to these init functions, although I see they do more than
> just trivial initialization, so keeping context analysis enabled might
> catch real bugs.
Makes sense. I think grouping all unpublished guarded variable initializations
under a single context_unsafe(...) block is a better approach than wrapping
each initialization separately, which can quickly clutter the code.
In fact, there are a few initialization functions in the NVMe driver where
I annotated the entire function with __context_unsafe(). I agree it would be
better to narrow the scope and annotate only the unpublished guarded variable
initializations using a single context_unsafe(...) block.
I'll update the next revision of the patchset accordingly and gather feedback
on that approach. Let's see what others think.
Thanks,
--Nilay
next prev parent reply other threads:[~2026-06-30 9:54 UTC|newest]
Thread overview: 68+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-06-14 13:15 [PATCHv2 00/17] Support Clang context analysis for NVMe host drivers Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 01/17] nvme: update nvme_passthru_end() signature Nilay Shroff
2026-06-26 6:32 ` Christoph Hellwig
2026-06-14 13:15 ` [PATCHv2 02/17] nvme: add Clang context annotations for nvme_passthru_{start|stop} Nilay Shroff
2026-06-26 6:33 ` Christoph Hellwig
2026-06-26 14:20 ` Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 03/17] nvme: add Clang context annotations for nvme_ns_head::srcu Nilay Shroff
2026-06-26 6:34 ` Christoph Hellwig
2026-06-14 13:15 ` [PATCHv2 04/17] nvme: add Clang context annotations for nvme_ns_head::requeue_list Nilay Shroff
2026-06-26 6:37 ` Christoph Hellwig
2026-06-26 14:24 ` Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 05/17] nvme: add Clang context annotations for nvme_ns_head::current_path Nilay Shroff
2026-06-26 6:40 ` Christoph Hellwig
2026-06-26 15:35 ` Nilay Shroff
2026-06-26 18:36 ` Paul E. McKenney
2026-06-27 15:38 ` Marco Elver
2026-06-27 17:14 ` Paul E. McKenney
2026-06-28 6:00 ` Marco Elver
2026-06-28 16:07 ` Paul E. McKenney
2026-06-29 5:20 ` Nilay Shroff
2026-06-29 8:48 ` Marco Elver
2026-06-29 15:33 ` Paul E. McKenney
2026-06-30 9:56 ` Nilay Shroff
2026-06-30 14:24 ` Paul E. McKenney
2026-06-29 6:00 ` Marco Elver
2026-06-29 6:04 ` Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 06/17] nvme: add Clang context annotations for nvme_dev::shutdown_lock Nilay Shroff
2026-06-26 6:41 ` Christoph Hellwig
2026-06-14 13:15 ` [PATCHv2 07/17] nvme: add Clang context annotations for nvme_subsystem::lock Nilay Shroff
2026-06-26 6:43 ` Christoph Hellwig
2026-06-26 14:39 ` Nilay Shroff
2026-06-29 12:47 ` Christoph Hellwig
2026-06-29 23:12 ` Marco Elver
2026-06-30 9:53 ` Nilay Shroff [this message]
2026-07-09 6:20 ` Christoph Hellwig
2026-07-09 7:24 ` Marco Elver
2026-07-10 7:58 ` Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 08/17] nvme: add Clang context annotations for nvme_ctrl::ana_lock Nilay Shroff
2026-06-26 6:43 ` Christoph Hellwig
2026-06-14 13:15 ` [PATCHv2 09/17] nvme: add Clang context annotations for nvme_subsystems_lock Nilay Shroff
2026-06-26 6:45 ` Christoph Hellwig
2026-06-26 14:48 ` Nilay Shroff
2026-06-29 12:47 ` Christoph Hellwig
2026-06-14 13:15 ` [PATCHv2 10/17] nvme: add Clang context annotations in fabric.c Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 11/17] nvme: add Clang context annotations for nvme_queue::sq_lock Nilay Shroff
2026-06-26 6:47 ` Christoph Hellwig
2026-06-26 9:50 ` Marco Elver
2026-06-26 15:12 ` Nilay Shroff
2026-06-29 12:48 ` Christoph Hellwig
2026-06-14 13:15 ` [PATCHv2 12/17] nvme: add Clang context annotations for nvme_queue::cq_poll_lock Nilay Shroff
2026-06-26 6:48 ` Christoph Hellwig
2026-06-26 15:22 ` Nilay Shroff
2026-06-29 12:50 ` Christoph Hellwig
2026-06-30 9:26 ` Nilay Shroff
2026-06-30 10:56 ` Marco Elver
2026-07-01 7:45 ` Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 13/17] nvme: add Clang context annotations in rdma.c Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 14/17] nvme: fix Clang context analysis warning " Nilay Shroff
2026-06-26 6:49 ` Christoph Hellwig
2026-06-26 15:31 ` Nilay Shroff
2026-06-29 12:50 ` Christoph Hellwig
2026-06-29 22:47 ` Marco Elver
2026-06-30 9:35 ` Nilay Shroff
2026-06-30 10:38 ` Marco Elver
2026-07-01 7:50 ` Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 15/17] nvme: add Clang context annotations in tcp.c Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 16/17] nvme: fix Clang context analysis warning " Nilay Shroff
2026-06-14 13:15 ` [PATCHv2 17/17] nvme: enable Clang context analysis support for nvme host driver Nilay Shroff
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3d7fd5f4-26b1-42a6-89c5-961ca53c51b1@linux.ibm.com \
--to=nilay@linux.ibm.com \
--cc=axboe@fb.com \
--cc=bvanassche@acm.org \
--cc=elver@google.com \
--cc=gjoyce@linux.ibm.com \
--cc=hch@lst.de \
--cc=kbusch@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nvme@lists.infradead.org \
--cc=sagi@grimberg.me \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®