From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jens Axboe <jaxboe-5c4llco8/ftWk0Htik3J/w@public.gmane.org>
Subject: Re: [Bug #32982] Kernel locks up a few minutes after boot
Date: Tue, 19 Apr 2011 19:06:18 +0200
Message-ID: <4DADC10A.3060809@fusionio.com>
References: <OlQwC5HnRsN.A.ULB.OGuqNB@chimera> <_H4l51C1wXN.A.yDC.yGuqNB@chimera> <BANLkTimeaHwBiZoz=isCM=qtBxoxaBqDUw@mail.gmail.com> <BANLkTi=1jO73DXDTD50bVpU82qphVKmRCw@mail.gmail.com> <4DAC2429.5000105@fusionio.com> <BANLkTinCWWYqnksxevho9TN3u6ZC_Rjzwg@mail.gmail.com> <4DAC82E6.3020809@fusionio.com> <BANLkTimaekpwmK+z=ePE+XSP6GUteidwfw@mail.gmail.com> <4DAD5156.2050300@fusionio.com> <4DAD6EF2.5070405@fusionio.com> <20110419164815.GA30616@infradead.org>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Return-path: <kernel-testers-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>
In-Reply-To: <20110419164815.GA30616-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
Sender: kernel-testers-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
List-ID: <kernel-testers.vger.kernel.org>
Content-Type: text/plain; charset="us-ascii"
To: Christoph Hellwig <hch-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
Cc: Bart Van Assche <bvanassche-HInyCGIudOg@public.gmane.org>, Linus Torvalds <torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>, "Rafael J. Wysocki" <rjw-KKrjLPT3xs0@public.gmane.org>, Linux Kernel Mailing List <linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>, Kernel Testers List <kernel-testers-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>, Maciej Rutecki <maciej.rutecki-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>, Florian Mickler <florian-sVu6HhrpSfRAfugRpC6u6w@public.gmane.org>, Neil Brown <neilb-l3A5Bk7waGM@public.gmane.org>

On 2011-04-19 18:48, Christoph Hellwig wrote:
>> +		blk_run_queue_async(sdev->request_queue);
> 
> This doesn't even have to be async except when scsi drivers call
> cmd->scsi_done directly.  It seems like if this always went through the
> softirq (or kblockd) we could still run it in context for the others.

Exactly. I'll pass an 'optimize' patch past James.

>> +	/*
>> +	 * This get/put dance makes no sense
>> +	 */
>>  	get_device(&rport->dev);
>> -
>> -	spin_lock_irqsave(rport->rqst_q->queue_lock, flags);
>> -	flagset = test_bit(QUEUE_FLAG_REENTER, &rport->rqst_q->queue_flags) &&
>> -		  !test_bit(QUEUE_FLAG_REENTER, &rport->rqst_q->queue_flags);
>> -	if (flagset)
>> -		queue_flag_set(QUEUE_FLAG_REENTER, rport->rqst_q);
>> -	__blk_run_queue(rport->rqst_q);
>> -	if (flagset)
>> -		queue_flag_clear(QUEUE_FLAG_REENTER, rport->rqst_q);
>> -	spin_unlock_irqrestore(rport->rqst_q->queue_lock, flags);
>> -
>> +	blk_run_queue_async(rport->rqst_q);
> 
> And the QUEUE_FLAG_REENTER mess here never made sense either as it
> tested for a bit beeing set and not set at the same time.  So this one
> actually should be able to be replaced by a plain blk_run_queue.

Yep, it's completely broken as-is.

-- 
Jens Axboe