All of lore.kernel.org
 help / color / mirror / Atom feed
From: Tony Battersby <tonyb@cybernetics.com>
To: Douglas Gilbert <dgilbert@interlog.com>,
	FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>,
	James.Bottomley@HansenPartnership.com
Cc: Christoph Hellwig <hch@infradead.org>, linux-scsi@vger.kernel.org
Subject: [PATCH 2/2] sg: fix races with ioctl(SG_IO)
Date: Mon, 05 Jan 2009 14:11:23 -0500	[thread overview]
Message-ID: <49625B5B.6090802@cybernetics.com> (raw)

sg_io_owned needs to be set before the command is sent to the midlevel;
otherwise, a quickly-completing command may cause a different CPU
to see "srp->done == 1 && !srp->sg_io_owned", which would lead to
incorrect behavior.

Check srp->done and set srp->orphan while holding rq_list_lock to
prevent races with sg_rq_end_io().

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
---
 sg.c |   21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)
--- linux-2.6.28/drivers/scsi/sg.c.orig	2009-01-05 11:08:09.000000000 -0500
+++ linux-2.6.28/drivers/scsi/sg.c	2009-01-05 11:10:46.000000000 -0500
@@ -197,7 +197,7 @@ static ssize_t sg_new_read(Sg_fd * sfp, 
 			   Sg_request * srp);
 static ssize_t sg_new_write(Sg_fd *sfp, struct file *file,
 			const char __user *buf, size_t count, int blocking,
-			int read_only, Sg_request **o_srp);
+			int read_only, int sg_io_owned, Sg_request **o_srp);
 static int sg_common_write(Sg_fd * sfp, Sg_request * srp,
 			   unsigned char *cmnd, int timeout, int blocking);
 static int sg_read_oxfer(Sg_request * srp, char __user *outp, int num_read_xfer);
@@ -607,7 +607,8 @@ sg_write(struct file *filp, const char _
 		return -EFAULT;
 	blocking = !(filp->f_flags & O_NONBLOCK);
 	if (old_hdr.reply_len < 0)
-		return sg_new_write(sfp, filp, buf, count, blocking, 0, NULL);
+		return sg_new_write(sfp, filp, buf, count,
+				    blocking, 0, 0, NULL);
 	if (count < (SZ_SG_HEADER + 6))
 		return -EIO;	/* The minimum scsi command length is 6 bytes. */
 
@@ -688,7 +689,7 @@ sg_write(struct file *filp, const char _
 
 static ssize_t
 sg_new_write(Sg_fd *sfp, struct file *file, const char __user *buf,
-		 size_t count, int blocking, int read_only,
+		 size_t count, int blocking, int read_only, int sg_io_owned,
 		 Sg_request **o_srp)
 {
 	int k;
@@ -708,6 +709,7 @@ sg_new_write(Sg_fd *sfp, struct file *fi
 		SCSI_LOG_TIMEOUT(1, printk("sg_new_write: queue full\n"));
 		return -EDOM;
 	}
+	srp->sg_io_owned = sg_io_owned;
 	hp = &srp->header;
 	if (__copy_from_user(hp, buf, SZ_SG_IO_HDR)) {
 		sg_remove_request(sfp, srp);
@@ -854,10 +856,9 @@ sg_ioctl(struct inode *inode, struct fil
 				return -EFAULT;
 			result =
 			    sg_new_write(sfp, filp, p, SZ_SG_IO_HDR,
-					 blocking, read_only, &srp);
+					 blocking, read_only, 1, &srp);
 			if (result < 0)
 				return result;
-			srp->sg_io_owned = 1;
 			while (1) {
 				result = 0;	/* following macro to beat race condition */
 				__wait_event_interruptible(sfp->read_wait,
@@ -867,14 +868,16 @@ sg_ioctl(struct inode *inode, struct fil
 					return -ENODEV;
 				if (sfp->closed)
 					return 0;	/* request packet dropped already */
-				if (0 == result)
+				write_lock_irq(&sfp->rq_list_lock);
+				if (srp->done) {
+					srp->done = 2;
+					write_unlock_irq(&sfp->rq_list_lock);
 					break;
+				}
 				srp->orphan = 1;
+				write_unlock_irq(&sfp->rq_list_lock);
 				return result;	/* -ERESTARTSYS because signal hit process */
 			}
-			write_lock_irqsave(&sfp->rq_list_lock, iflags);
-			srp->done = 2;
-			write_unlock_irqrestore(&sfp->rq_list_lock, iflags);
 			result = sg_new_read(sfp, p, SZ_SG_IO_HDR, srp);
 			return (result < 0) ? result : 0;
 		}



             reply	other threads:[~2009-01-05 19:11 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-01-05 19:11 Tony Battersby [this message]
  -- strict thread matches above, loose matches on Subject: below --
2009-01-05 19:07 [PATCH 0/2] sg: fix races during device removal (v2) Tony Battersby
2009-01-10 17:26 ` FUJITA Tomonori
2009-01-14 20:31   ` Tony Battersby
2009-01-19  6:57     ` FUJITA Tomonori
2009-01-19 23:06       ` [PATCH 2/2] sg: fix races with ioctl(SG_IO) Tony Battersby

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=49625B5B.6090802@cybernetics.com \
    --to=tonyb@cybernetics.com \
    --cc=James.Bottomley@HansenPartnership.com \
    --cc=dgilbert@interlog.com \
    --cc=fujita.tomonori@lab.ntt.co.jp \
    --cc=hch@infradead.org \
    --cc=linux-scsi@vger.kernel.org \
    /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.