linux-block.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Michael Schmitz <schmitzmic@gmail.com>
To: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>,
	linux-block@vger.kernel.org, linux-m68k@lists.linux-m68k.org,
	Christoph Hellwig <hch@lst.de>,
	Luis Chamberlain <mcgrof@kernel.org>,
	Finn Thain <fthain@linux-m68k.org>
Subject: Re: [PATCH v2] ataflop: remove ataflop_probe_lock mutex
Date: Mon, 18 Oct 2021 08:05:28 +1300	[thread overview]
Message-ID: <7bf8ba1a-ed7c-059b-f4a1-cd7cfb5c21da@gmail.com> (raw)
In-Reply-To: <5524e6ee-e469-9775-07c4-7baf5e330148@i-love.sakura.ne.jp>

Hi Tetsuo,

On 17/10/21 15:09, Tetsuo Handa wrote:
> Commit bf9c0538e485b591 ("ataflop: use a separate gendisk for each media
> format") introduced ataflop_probe_lock mutex, but forgot to unlock the
> mutex when atari_floppy_init() (i.e. module loading) succeeded. This will
> result in double lock deadlock if ataflop_probe() is called. Also,
> unregister_blkdev() must not be called from atari_floppy_init() with
> ataflop_probe_lock held when atari_floppy_init() failed, for
> ataflop_probe() waits for ataflop_probe_lock with major_names_lock held
> (i.e. AB-BA deadlock).
>
> __register_blkdev() needs to be called last in order to avoid calling
> ataflop_probe() when atari_floppy_init() is about to fail, for memory for
> completing already-started ataflop_probe() safely will be released as soon
> as atari_floppy_init() released ataflop_probe_lock mutex.
>
> As with commit 8b52d8be86d72308 ("loop: reorder loop_exit"),
> unregister_blkdev() needs to be called first in order to avoid calling
> ataflop_alloc_disk() from ataflop_probe() after del_gendisk() from
> atari_floppy_exit().
>
> By relocating __register_blkdev() / unregister_blkdev() as explained above,
> we can remove ataflop_probe_lock mutex, for probe function and __exit
> function are serialized by major_names_lock mutex.
>
> Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> Fixes: bf9c0538e485b591 ("ataflop: use a separate gendisk for each media format")
> ---
> Changes in v2:
>   Remove ataflop_probe_lock mutex than unlocking.
>
> Finn Thain wrote:
>> So I wonder if it would have been possible to use Aranym to find the
>> regression, or avoid it in the first place?
>
> OK, there is an emulator for testing this module. But I'm not familiar
> with m68k environment. Luis Chamberlain is proposing patchset for adding
> add_disk() error handling. I think that an answer would be to include
> m68k's mailing list into a patch for this module in order to notify of
> changes and expect m68k developers to review/test the patch.
>
> Michael Schmitz wrote:
>> Not as a module, no. I use the Atari floppy driver built-in. Latest kernel version I ran was 5.13.
>
> Great. Can you try this patch alone?

Doesn't appear to work, sorry.

Regards,

	Michael Schmitz

>
>  drivers/block/ataflop.c | 55 ++++++++++++++++++++---------------------
>  1 file changed, 27 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/block/ataflop.c b/drivers/block/ataflop.c
> index a093644ac39f..adfe198e4699 100644
> --- a/drivers/block/ataflop.c
> +++ b/drivers/block/ataflop.c
> @@ -1986,8 +1986,6 @@ static int ataflop_alloc_disk(unsigned int drive, unsigned int type)
>  	return 0;
>  }
>
> -static DEFINE_MUTEX(ataflop_probe_lock);
> -
>  static void ataflop_probe(dev_t dev)
>  {
>  	int drive = MINOR(dev) & 3;
> @@ -1998,12 +1996,30 @@ static void ataflop_probe(dev_t dev)
>
>  	if (drive >= FD_MAX_UNITS || type >= NUM_DISK_MINORS)
>  		return;
> -	mutex_lock(&ataflop_probe_lock);
>  	if (!unit[drive].disk[type]) {
>  		if (ataflop_alloc_disk(drive, type) == 0)
>  			add_disk(unit[drive].disk[type]);
>  	}
> -	mutex_unlock(&ataflop_probe_lock);
> +}
> +
> +static void atari_floppy_cleanup(void)
> +{
> +	int i;
> +	int type;
> +
> +	for (i = 0; i < FD_MAX_UNITS; i++) {
> +		for (type = 0; type < NUM_DISK_MINORS; type++) {
> +			if (!unit[i].disk[type])
> +				continue;
> +			del_gendisk(unit[i].disk[type]);
> +			blk_cleanup_queue(unit[i].disk[type]->queue);
> +			put_disk(unit[i].disk[type]);
> +		}
> +		blk_mq_free_tag_set(&unit[i].tag_set);
> +	}
> +
> +	del_timer_sync(&fd_timer);
> +	atari_stram_free(DMABuffer);
>  }
>
>  static int __init atari_floppy_init (void)
> @@ -2015,11 +2031,6 @@ static int __init atari_floppy_init (void)
>  		/* Amiga, Mac, ... don't have Atari-compatible floppy :-) */
>  		return -ENODEV;
>
> -	mutex_lock(&ataflop_probe_lock);
> -	ret = __register_blkdev(FLOPPY_MAJOR, "fd", ataflop_probe);
> -	if (ret)
> -		goto out_unlock;
> -
>  	for (i = 0; i < FD_MAX_UNITS; i++) {
>  		memset(&unit[i].tag_set, 0, sizeof(unit[i].tag_set));
>  		unit[i].tag_set.ops = &ataflop_mq_ops;
> @@ -2072,7 +2083,12 @@ static int __init atari_floppy_init (void)
>  	       UseTrackbuffer ? "" : "no ");
>  	config_types();
>
> -	return 0;
> +	ret = __register_blkdev(FLOPPY_MAJOR, "fd", ataflop_probe);
> +	if (ret) {
> +		printk(KERN_ERR "atari_floppy_init: cannot register block device\n");
> +		atari_floppy_cleanup();
> +	}
> +	return ret;
>
>  err:
>  	while (--i >= 0) {
> @@ -2081,9 +2097,6 @@ static int __init atari_floppy_init (void)
>  		blk_mq_free_tag_set(&unit[i].tag_set);
>  	}
>
> -	unregister_blkdev(FLOPPY_MAJOR, "fd");
> -out_unlock:
> -	mutex_unlock(&ataflop_probe_lock);
>  	return ret;
>  }
>
> @@ -2128,22 +2141,8 @@ __setup("floppy=", atari_floppy_setup);
>
>  static void __exit atari_floppy_exit(void)
>  {
> -	int i, type;
> -
> -	for (i = 0; i < FD_MAX_UNITS; i++) {
> -		for (type = 0; type < NUM_DISK_MINORS; type++) {
> -			if (!unit[i].disk[type])
> -				continue;
> -			del_gendisk(unit[i].disk[type]);
> -			blk_cleanup_queue(unit[i].disk[type]->queue);
> -			put_disk(unit[i].disk[type]);
> -		}
> -		blk_mq_free_tag_set(&unit[i].tag_set);
> -	}
>  	unregister_blkdev(FLOPPY_MAJOR, "fd");
> -
> -	del_timer_sync(&fd_timer);
> -	atari_stram_free( DMABuffer );
> +	atari_floppy_cleanup();
>  }
>
>  module_init(atari_floppy_init)
>

  reply	other threads:[~2021-10-17 19:05 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-10-16 13:25 [PATCH] ataflop: unlock ataflop_probe_lock at atari_floppy_init() Tetsuo Handa
2021-10-16 22:56 ` Finn Thain
2021-10-17  1:52 ` Michael Schmitz
2021-10-17  2:09   ` [PATCH v2] ataflop: remove ataflop_probe_lock mutex Tetsuo Handa
2021-10-17 19:05     ` Michael Schmitz [this message]
2021-10-17 23:47       ` Michael Schmitz
2021-10-18  8:15         ` Michael Schmitz
2021-10-18 22:25     ` Michael Schmitz
2021-10-21 16:20     ` Luis Chamberlain
2021-10-21 16:21       ` Luis Chamberlain

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=7bf8ba1a-ed7c-059b-f4a1-cd7cfb5c21da@gmail.com \
    --to=schmitzmic@gmail.com \
    --cc=fthain@linux-m68k.org \
    --cc=hch@lst.de \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-m68k@lists.linux-m68k.org \
    --cc=mcgrof@kernel.org \
    --cc=penguin-kernel@i-love.sakura.ne.jp \
    /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).