Linux-SCSI Archive on lore.kernel.org
 help / color / Atom feed
* [PATCH v3] scsi: core: fix uninit-value access of variable sshdr
@ 2019-10-12  3:25 zhengbin
  2019-10-17  0:05 ` Finn Thain
  2019-10-17  2:45 ` Bart Van Assche
  0 siblings, 2 replies; 5+ messages in thread
From: zhengbin @ 2019-10-12  3:25 UTC (permalink / raw)
  To: bvanassche, jejb, martin.petersen, linux-scsi; +Cc: yi.zhang, zhengbin13

kmsan report a warning in 5.1-rc4:

BUG: KMSAN: uninit-value in sr_get_events drivers/scsi/sr.c:207 [inline]
BUG: KMSAN: uninit-value in sr_check_events+0x2cf/0x1090 drivers/scsi/sr.c:243
CPU: 1 PID: 13858 Comm: syz-executor.0 Tainted: G    B             5.1.0-rc4+ #8
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x173/0x1d0 lib/dump_stack.c:113
 kmsan_report+0x131/0x2a0 mm/kmsan/kmsan.c:619
 __msan_warning+0x7a/0xf0 mm/kmsan/kmsan_instr.c:310
 sr_get_events drivers/scsi/sr.c:207 [inline]
 sr_check_events+0x2cf/0x1090 drivers/scsi/sr.c:243

The reason is as follows:
sr_get_events
  struct scsi_sense_hdr sshdr;  -->uninit
  scsi_execute_req              -->If fail, will not set sshdr
  scsi_sense_valid(&sshdr)      -->access sshdr

We can init sshdr in sr_get_events, but there have many callers of
scsi_execute, scsi_execute_req, we have to troubleshoot all callers,
the simpler way is init sshdr in __scsi_execute.

BTW: sr_do_ioctl should check whether sshdr is valid, fix this

Signed-off-by: zhengbin <zhengbin13@huawei.com>
---
v1->v2: modify the comments suggested by Bart
v2->v3: fix bug in sr_do_ioctl
 drivers/scsi/scsi_lib.c | 7 +++++++
 drivers/scsi/sr_ioctl.c | 5 +++++
 2 files changed, 12 insertions(+)

diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 5447738..d5e29c5 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -255,6 +255,13 @@ int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
 	struct scsi_request *rq;
 	int ret = DRIVER_ERROR << 24;

+	/*
+	 * Zero-initialize sshdr for those callers that check the *sshdr
+	 * contents even if no sense data is available.
+	 */
+	if (sshdr)
+		memset(sshdr, 0, sizeof(struct scsi_sense_hdr));
+
 	req = blk_get_request(sdev->request_queue,
 			data_direction == DMA_TO_DEVICE ?
 			REQ_OP_SCSI_OUT : REQ_OP_SCSI_IN, BLK_MQ_REQ_PREEMPT);
diff --git a/drivers/scsi/sr_ioctl.c b/drivers/scsi/sr_ioctl.c
index ffcf902..335cfdd 100644
--- a/drivers/scsi/sr_ioctl.c
+++ b/drivers/scsi/sr_ioctl.c
@@ -206,6 +206,11 @@ int sr_do_ioctl(Scsi_CD *cd, struct packet_command *cgc)

 	/* Minimal error checking.  Ignore cases we know about, and report the rest. */
 	if (driver_byte(result) != 0) {
+		if (!scsi_sense_valid(sshdr)) {
+			err = -EIO;
+			goto out;
+		}
+
 		switch (sshdr->sense_key) {
 		case UNIT_ATTENTION:
 			SDev->changed = 1;
--
2.7.4


^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, back to index

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-10-12  3:25 [PATCH v3] scsi: core: fix uninit-value access of variable sshdr zhengbin
2019-10-17  0:05 ` Finn Thain
2019-10-17  1:16   ` zhengbin (A)
2019-10-17  2:45 ` Bart Van Assche
2019-10-17  3:26   ` zhengbin (A)

Linux-SCSI Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-scsi/0 linux-scsi/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-scsi linux-scsi/ https://lore.kernel.org/linux-scsi \
		linux-scsi@vger.kernel.org
	public-inbox-index linux-scsi

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-scsi


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git