lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20180406083016.weuoemmcgvgxlcxy@pathway.suse.cz>
Date:   Fri, 6 Apr 2018 10:30:16 +0200
From:   Petr Mladek <pmladek@...e.com>
To:     wen.yang99@....com.cn
Cc:     sergey.senozhatsky.work@...il.com, jejb@...ux.vnet.ibm.com,
        martin.petersen@...cle.com, linux-scsi@...r.kernel.org,
        linux-kernel@...r.kernel.org, Bart.VanAssche@....com,
        tj@...nel.org, jiang.biao2@....com.cn, zhong.weidong@....com.cn,
        tan.hu@....com.cn
Subject: Re: 答复: Re: [PATCH v2] scsi: Introduce sdev_printk_ratelimited to
 throttlefrequent printk

On Tue 2018-04-03 14:19:43, wen.yang99@....com.cn wrote:
> On the other hand,queue_lock is big, looping doing something under spinlock 
> 
> may locked many things and taking a long time, may cause some problems.
> 
> So This code needs to be optimized later:
> 
> scsi_request_fn()
> {
> 	for (;;) {
> 		int rtn;
> 		/*
> 		 * get next queueable request.  We do this early to make sure
> 		 * that the request is fully prepared even if we cannot
> 		 * accept it.
> 		 */
> 
> 		req = blk_peek_request(q);
> 
> 		if (!req)
> 			break;
> 
> 		if (unlikely(!scsi_device_online(sdev))) {
> 			sdev_printk(KERN_ERR, sdev,
> 				    "rejected I/O to offline device\n");
> 			scsi_kill_request(req, q);
> 			continue;
> 
> 			^^^^^^^^^ still under spinlock
> 		}

I wonder if the following might be the best solution after all:

		if (unlikely(!scsi_device_online(sdev))) {
			scsi_kill_request(req, q);

			/*
			 * printk() might take a while on slow consoles.
			 * Prevent solftlockups by releasing the lock.
			 */
			spin_unlock_irq(q->queue_lock);
			sdev_printk(KERN_ERR, sdev,
				    "rejecting I/O to offline device\n");
			spin_lock_irq(q->queue_lock);
			continue;
		}

I see that the lock is released also in several other situations.
Therefore it looks safe. Also handling too many requests without
releasing the lock seems to be a bad idea in general. I think
that this solution was already suggested earlier.

Please, note that I moved scsi_kill_request() up. It looks natural
to remove it from the queue before we release the queue lock.

Best Regards,
Petr

BTW: Your mail had strange formatting. Please, try to avoid using
html.

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ