All of lore.kernel.org
 help / color / mirror / Atom feed
From: Javier Martinez Canillas <javierm@redhat.com>
To: Daniel Vetter <daniel.vetter@ffwll.ch>,
	DRI Development <dri-devel@lists.freedesktop.org>
Cc: Du Cheng <ducheng2@gmail.com>,
	Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Intel Graphics Development <intel-gfx@lists.freedesktop.org>,
	Zheyu Ma <zheyuma97@gmail.com>,
	Matthew Wilcox <willy@infradead.org>,
	Nathan Chancellor <nathan@kernel.org>,
	Claudio Suarez <cssk@net-c.es>,
	Thomas Zimmermann <tzimmermann@suse.de>,
	Daniel Vetter <daniel.vetter@intel.com>,
	Geert Uytterhoeven <geert@linux-m68k.org>,
	Sam Ravnborg <sam@ravnborg.org>, Helge Deller <deller@gmx.de>,
	Guenter Roeck <linux@roeck-us.net>
Subject: Re: [PATCH] fbcon: Fix delayed takeover locking
Date: Wed, 13 Apr 2022 11:16:08 +0200	[thread overview]
Message-ID: <955400a5-a785-57af-92d4-42710c55ce99@redhat.com> (raw)
In-Reply-To: <20220413082128.348186-1-daniel.vetter@ffwll.ch>

Hello Daniel,

On 4/13/22 10:21, Daniel Vetter wrote:
> I messed up the delayed takover path in the locking conversion in
> 6e7da3af008b ("fbcon: Move console_lock for register/unlink/unregister").
>

Maybe a few more words of what the issue is ? Something like the following:

If CONFIG_FRAMEBUFFER_CONSOLE_DEFERRED_TAKEOVER is enabled, fbcon take-over
doesn't take place when calling fbcon_fb_registered(). Instead, is deferred
using a workqueue and its fbcon_register_existing_fbs() function calls to
fbcon_fb_registered() again for each registered fbcon fb.

This leads to the console_lock tried to be held twice, causing a deadlock.
 
> Fix it by re-extracting the lockless function and using it in the
> delayed takeover path, where we need to hold the lock already to
> iterate over the list of already registered fb. Well the current code
> still is broken in there (since the list is protected by a
> registration_lock, which we can't take here because it nests the other
> way round with console_lock), but in the future this will be a list
> protected by console_lock when this is all sorted out.
> 

[snip]

>  
> -/* called with console_lock held */
>  void fbcon_fb_unbind(struct fb_info *info)
>  {
>  	int i, new_idx = -1;
> @@ -2822,7 +2821,6 @@ void fbcon_fb_unbind(struct fb_info *info)
>  	console_unlock();
>  }
>  
> -/* called with console_lock held */
>  void fbcon_fb_unregistered(struct fb_info *info)
>  {

Removing these comments feels like should be in a separate patch or at least
mention in the patch description that should had been removed in the commit
6e7da3af008b ("fbcon: Move console_lock for register/unlink/unregister"),
that made these functions to be called without the console_lock being held.

The changes themselves look good to me.

Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>

-- 
Best regards,

Javier Martinez Canillas
Linux Engineering
Red Hat


WARNING: multiple messages have this Message-ID (diff)
From: Javier Martinez Canillas <javierm@redhat.com>
To: Daniel Vetter <daniel.vetter@ffwll.ch>,
	DRI Development <dri-devel@lists.freedesktop.org>
Cc: Du Cheng <ducheng2@gmail.com>,
	Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Intel Graphics Development <intel-gfx@lists.freedesktop.org>,
	Zheyu Ma <zheyuma97@gmail.com>,
	Matthew Wilcox <willy@infradead.org>,
	Nathan Chancellor <nathan@kernel.org>,
	Thomas Zimmermann <tzimmermann@suse.de>,
	Daniel Vetter <daniel.vetter@intel.com>,
	Geert Uytterhoeven <geert@linux-m68k.org>,
	Sam Ravnborg <sam@ravnborg.org>, Helge Deller <deller@gmx.de>,
	Guenter Roeck <linux@roeck-us.net>
Subject: Re: [Intel-gfx] [PATCH] fbcon: Fix delayed takeover locking
Date: Wed, 13 Apr 2022 11:16:08 +0200	[thread overview]
Message-ID: <955400a5-a785-57af-92d4-42710c55ce99@redhat.com> (raw)
In-Reply-To: <20220413082128.348186-1-daniel.vetter@ffwll.ch>

Hello Daniel,

On 4/13/22 10:21, Daniel Vetter wrote:
> I messed up the delayed takover path in the locking conversion in
> 6e7da3af008b ("fbcon: Move console_lock for register/unlink/unregister").
>

Maybe a few more words of what the issue is ? Something like the following:

If CONFIG_FRAMEBUFFER_CONSOLE_DEFERRED_TAKEOVER is enabled, fbcon take-over
doesn't take place when calling fbcon_fb_registered(). Instead, is deferred
using a workqueue and its fbcon_register_existing_fbs() function calls to
fbcon_fb_registered() again for each registered fbcon fb.

This leads to the console_lock tried to be held twice, causing a deadlock.
 
> Fix it by re-extracting the lockless function and using it in the
> delayed takeover path, where we need to hold the lock already to
> iterate over the list of already registered fb. Well the current code
> still is broken in there (since the list is protected by a
> registration_lock, which we can't take here because it nests the other
> way round with console_lock), but in the future this will be a list
> protected by console_lock when this is all sorted out.
> 

[snip]

>  
> -/* called with console_lock held */
>  void fbcon_fb_unbind(struct fb_info *info)
>  {
>  	int i, new_idx = -1;
> @@ -2822,7 +2821,6 @@ void fbcon_fb_unbind(struct fb_info *info)
>  	console_unlock();
>  }
>  
> -/* called with console_lock held */
>  void fbcon_fb_unregistered(struct fb_info *info)
>  {

Removing these comments feels like should be in a separate patch or at least
mention in the patch description that should had been removed in the commit
6e7da3af008b ("fbcon: Move console_lock for register/unlink/unregister"),
that made these functions to be called without the console_lock being held.

The changes themselves look good to me.

Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>

-- 
Best regards,

Javier Martinez Canillas
Linux Engineering
Red Hat


  reply	other threads:[~2022-04-13  9:16 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-04-13  8:21 [PATCH] fbcon: Fix delayed takeover locking Daniel Vetter
2022-04-13  8:21 ` [Intel-gfx] " Daniel Vetter
2022-04-13  9:16 ` Javier Martinez Canillas [this message]
2022-04-13  9:16   ` Javier Martinez Canillas
2022-04-13  9:21   ` Daniel Vetter
2022-04-13  9:21     ` [Intel-gfx] " Daniel Vetter
2022-04-13  9:31     ` Javier Martinez Canillas
2022-04-13  9:31       ` [Intel-gfx] " Javier Martinez Canillas
2022-04-13  9:25 ` Daniel Vetter
2022-04-13  9:25   ` [Intel-gfx] " Daniel Vetter
2022-04-13 15:20   ` Nathan Chancellor
2022-04-13 15:20     ` [Intel-gfx] " Nathan Chancellor
2022-04-13 20:50 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for " Patchwork
2022-04-13 21:12 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2022-04-13 23:13 ` [Intel-gfx] ✗ Fi.CI.IGT: failure " Patchwork

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=955400a5-a785-57af-92d4-42710c55ce99@redhat.com \
    --to=javierm@redhat.com \
    --cc=cssk@net-c.es \
    --cc=daniel.vetter@ffwll.ch \
    --cc=daniel.vetter@intel.com \
    --cc=deller@gmx.de \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=ducheng2@gmail.com \
    --cc=geert@linux-m68k.org \
    --cc=gregkh@linuxfoundation.org \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=linux@roeck-us.net \
    --cc=nathan@kernel.org \
    --cc=penguin-kernel@I-love.SAKURA.ne.jp \
    --cc=sam@ravnborg.org \
    --cc=tzimmermann@suse.de \
    --cc=willy@infradead.org \
    --cc=zheyuma97@gmail.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.