| Message ID | 20220602181818.50729-5-logang@deltatee.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Bug fixes for mdadm tests | expand |
On Thu, Jun 02, 2022 at 12:18:10PM -0600, Logan Gunthorpe wrote: > conf = mddev->private; > if (!conf || !conf->log) { > - spin_unlock(&mddev->lock); > + mddev_unlock(mddev); > return 0; > } > > @@ -2557,7 +2560,7 @@ static ssize_t r5c_journal_mode_show(struct mddev *mddev, char *page) > default: > ret = 0; > } > - spin_unlock(&mddev->lock); > + mddev_unlock(mddev); > return ret; Using a goto out_unlock would be nice here to keep the critical sections simple. But even as-is this looks good: Reviewed-by: Christoph Hellwig <hch@lst.de> > + lockdep_assert_held(&conf->mddev->reconfig_mutex); > + .. but this looks unrelated and misplaced in this patch.
On 2022-06-03 00:39, Christoph Hellwig wrote: > On Thu, Jun 02, 2022 at 12:18:10PM -0600, Logan Gunthorpe wrote: >> conf = mddev->private; >> if (!conf || !conf->log) { >> - spin_unlock(&mddev->lock); >> + mddev_unlock(mddev); >> return 0; >> } >> >> @@ -2557,7 +2560,7 @@ static ssize_t r5c_journal_mode_show(struct mddev *mddev, char *page) >> default: >> ret = 0; >> } >> - spin_unlock(&mddev->lock); >> + mddev_unlock(mddev); >> return ret; > > Using a goto out_unlock would be nice here to keep the critical > sections simple. But even as-is this looks good: > > Reviewed-by: Christoph Hellwig <hch@lst.de> > >> + lockdep_assert_held(&conf->mddev->reconfig_mutex); >> + > > .. but this looks unrelated and misplaced in this patch. > Ok, yup, I'll fix these issues up and send an updated series next week. Logan
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c index 83c184eddbda..69b95005abca 100644 --- a/drivers/md/raid5-cache.c +++ b/drivers/md/raid5-cache.c @@ -2534,10 +2534,13 @@ static ssize_t r5c_journal_mode_show(struct mddev *mddev, char *page) struct r5conf *conf; int ret; - spin_lock(&mddev->lock); + ret = mddev_lock(mddev); + if (ret) + return ret; + conf = mddev->private; if (!conf || !conf->log) { - spin_unlock(&mddev->lock); + mddev_unlock(mddev); return 0; } @@ -2557,7 +2560,7 @@ static ssize_t r5c_journal_mode_show(struct mddev *mddev, char *page) default: ret = 0; } - spin_unlock(&mddev->lock); + mddev_unlock(mddev); return ret; } @@ -3167,6 +3170,8 @@ void r5l_exit_log(struct r5conf *conf) { struct r5l_log *log = conf->log; + lockdep_assert_held(&conf->mddev->reconfig_mutex); + conf->log = NULL; synchronize_rcu();
The mddev->lock spinlock doesn't protect against the removal of conf->log in r5l_exit_log() so conf->log may be freed before it is used. To fix this, take the mddev_lock() insteaad of the mddev->lock spinlock. Signed-off-by: Logan Gunthorpe <logang@deltatee.com> --- drivers/md/raid5-cache.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)