public inbox for linux-scsi@vger.kernel.org
 help / color / mirror / Atom feed
From: Al Viro <viro@ZenIV.linux.org.uk>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: linux-scsi@vger.kernel.org
Subject: [RFC] apparently broken error recovery in vhost_scsi_iov_to_sgl()
Date: Sun, 24 Sep 2017 23:36:33 +0100	[thread overview]
Message-ID: <20170924223633.GV32076@ZenIV.linux.org.uk> (raw)

Suppose vhost_scsi_iov_to_sgl() got a two-iovec array, mapped
e.g. 20 pages from the first one just fine and failed on the
second.

static int
vhost_scsi_iov_to_sgl(struct vhost_scsi_cmd *cmd, bool write,
                      struct iov_iter *iter,
                      struct scatterlist *sg, int sg_count)
{
        size_t off = iter->iov_offset;
        int i, ret;

        for (i = 0; i < iter->nr_segs; i++) {
                void __user *base = iter->iov[i].iov_base + off;
                size_t len = iter->iov[i].iov_len - off;

                ret = vhost_scsi_map_to_sgl(cmd, base, len, sg, write);
                if (ret < 0) {
                        for (i = 0; i < sg_count; i++) {
                                struct page *page = sg_page(&sg[i]);
                                if (page)
                                        put_page(page);
                        }
                        return ret;
                }
                sg += ret;
                off = 0;
        }
        return 0;
}

What are we trying to drop in the if (ret < 0) in there?  In the case
above we step into it on the second pass through the loop.  The first
20 entries of sg had been filled... and sg had been increased by 20,
so whatever we find and feed to put_page(), it won't be those 20 pages.
Moreover, the caller will reset cmd->tvc_{prot_,}sgl_count to zero,
so vhost_scsi_release_cmd() won't find them either.

Am I missing something subtle here, or should that thing be doing
something like

diff --git a/drivers/vhost/scsi.c b/drivers/vhost/scsi.c
index 046f6d280af5..e47c5bc3ddca 100644
--- a/drivers/vhost/scsi.c
+++ b/drivers/vhost/scsi.c
@@ -688,6 +688,7 @@ vhost_scsi_iov_to_sgl(struct vhost_scsi_cmd *cmd, bool write,
 		      struct scatterlist *sg, int sg_count)
 {
 	size_t off = iter->iov_offset;
+	struct scatterlist *p = sg;
 	int i, ret;
 
 	for (i = 0; i < iter->nr_segs; i++) {
@@ -696,8 +697,8 @@ vhost_scsi_iov_to_sgl(struct vhost_scsi_cmd *cmd, bool write,
 
 		ret = vhost_scsi_map_to_sgl(cmd, base, len, sg, write);
 		if (ret < 0) {
-			for (i = 0; i < sg_count; i++) {
-				struct page *page = sg_page(&sg[i]);
+			while (p < sg) {
+				struct page *page = sg_page(p++);
 				if (page)
 					put_page(page);
 			}

             reply	other threads:[~2017-09-24 22:36 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-09-24 22:36 Al Viro [this message]
2017-09-25 21:53 ` [RFC] apparently broken error recovery in vhost_scsi_iov_to_sgl() Michael S. Tsirkin

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=20170924223633.GV32076@ZenIV.linux.org.uk \
    --to=viro@zeniv.linux.org.uk \
    --cc=linux-scsi@vger.kernel.org \
    --cc=mst@redhat.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox