From: Eric Sandeen <sandeen@sandeen.net>
To: Pavel Reichl <preichl@redhat.com>, Christoph Hellwig <hch@infradead.org>
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>, linux-xfs@vger.kernel.org
Subject: Re: [PATCH v5 1/4] xfs: Refactor xfs_isilocked()
Date: Fri, 21 Feb 2020 14:28:10 -0600 [thread overview]
Message-ID: <febec3d8-297b-9f67-b113-a068199a84e4@sandeen.net> (raw)
In-Reply-To: <CAJc7PzWVnV+ny_13rZVjEq_GMYWQciH_hWm+OXkw-OFQtn-zDg@mail.gmail.com>
On 2/21/20 11:49 AM, Pavel Reichl wrote:
> On Wed, Feb 19, 2020 at 7:40 PM Christoph Hellwig <hch@infradead.org> wrote:
>>
>> On Tue, Feb 18, 2020 at 08:48:21PM -0800, Darrick J. Wong wrote:
>>>>> +static inline bool
>>>>> +__xfs_rwsem_islocked(
>>>>> + struct rw_semaphore *rwsem,
>>>>> + bool shared,
>>>>> + bool excl)
>>>>> +{
>>>>> + bool locked = false;
>>>>> +
>>>>> + if (!rwsem_is_locked(rwsem))
>>>>> + return false;
>>>>> +
>>>>> + if (!debug_locks)
>>>>> + return true;
>>>>> +
>>>>> + if (shared)
>>>>> + locked = lockdep_is_held_type(rwsem, 0);
>>>>> +
>>>>> + if (excl)
>>>>> + locked |= lockdep_is_held_type(rwsem, 1);
>>>>> +
>>>>> + return locked;
>>>>
>>>> This could use some comments explaining the logic, especially why we
>>>> need the shared and excl flags, which seems very confusing given that
>>>> a lock can be held either shared or exclusive, but not neither and not
>>>> both.
>>>
>>> Yes, this predicate should document that callers are allowed to pass in
>>> shared==true and excl==true when the caller wants to assert that either
>>> lock type (shared or excl) of a given lock class (e.g. iolock) are held.
>>
>> Looking at the lockdep_is_held_type implementation, and our existing
>> code for i_rwsem I really don't see the point of the extra shared
>> check. Something like:
>>
>> static inline bool
>> __xfs_rwsem_islocked(
>> struct rw_semaphore *rwsem,
>> bool excl)
>> {
>> if (rwsem_is_locked(rwsem)) {
>> if (debug_locks && excl)
>> return lockdep_is_held_type(rwsem, 1);
>> return true;
>> }
>>
>> return false;
>> }
>>
>> should be all that we really need.
>>
>
> You don't see the point of extra shared check, but if we want to check
> that the semaphore is locked for reading and not writing? Having the
> semaphore locked for writing would make the code safe from race
> condition but could be a performance hit, right?
So, I raised this question with Pavel but I think maybe it was borne
of my misunderstanding.
Ok let me think this through. Today we have:
int
xfs_isilocked(
xfs_inode_t *ip,
uint lock_flags)
{
if (lock_flags & (XFS_ILOCK_EXCL|XFS_ILOCK_SHARED)) {
if (!(lock_flags & XFS_ILOCK_SHARED))
return !!ip->i_lock.mr_writer;
return rwsem_is_locked(&ip->i_lock.mr_lock);
}
....
If we assert xfs_isilocked(ip, XFS_ILOCK_SHARED) I guess we /already/ get a positive
result if the inode is actually locked XFS_ILOCK_EXCL. So perhaps Christoph's
suggestion really just keeps implementing what we already have today.
It might be a reasonable question re: whether we ever want to know that we are locked
shared and NOT locked exclusive, but we can't do that today, so I guess it shouldn't
complicate this patchset.
... do I have this right?
Thanks,
-Eric
prev parent reply other threads:[~2020-02-21 20:28 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-14 18:59 [PATCH v5 1/4] xfs: Refactor xfs_isilocked() Pavel Reichl
2020-02-14 18:59 ` [PATCH v5 2/4] xfs: clean up whitespace in xfs_isilocked() calls Pavel Reichl
2020-02-16 22:36 ` Dave Chinner
2020-02-17 13:33 ` Christoph Hellwig
2020-02-14 18:59 ` [PATCH v5 3/4] xfs: xfs_isilocked() can only check a single lock type Pavel Reichl
2020-02-16 22:37 ` Dave Chinner
2020-02-17 13:34 ` Christoph Hellwig
2020-02-14 18:59 ` [PATCH v5 4/4] xfs: replace mrlock_t with rw_semaphores Pavel Reichl
2020-02-16 22:39 ` Dave Chinner
2020-02-17 13:35 ` Christoph Hellwig
2020-02-15 1:38 ` [PATCH v5 1/4] xfs: Refactor xfs_isilocked() Chaitanya Kulkarni
2020-02-17 10:55 ` Pavel Reichl
2020-02-20 16:25 ` Chaitanya Kulkarni
2020-02-16 22:36 ` Dave Chinner
2020-02-17 13:35 ` Christoph Hellwig
2020-02-19 4:48 ` Darrick J. Wong
2020-02-19 17:31 ` Pavel Reichl
2020-02-19 18:40 ` Christoph Hellwig
2020-02-19 20:16 ` Eric Sandeen
2020-02-20 16:30 ` Pavel Reichl
2020-02-20 16:32 ` Christoph Hellwig
2020-02-20 17:26 ` Eric Sandeen
2020-02-20 17:27 ` Darrick J. Wong
2020-02-21 17:49 ` Pavel Reichl
2020-02-21 20:28 ` Eric Sandeen [this message]
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=febec3d8-297b-9f67-b113-a068199a84e4@sandeen.net \
--to=sandeen@sandeen.net \
--cc=darrick.wong@oracle.com \
--cc=hch@infradead.org \
--cc=linux-xfs@vger.kernel.org \
--cc=preichl@redhat.com \
/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;
as well as URLs for NNTP newsgroup(s).