diff mbox

[v2,0/6] SCSI target patches for kernel v4.12

Message ID 1494523276.14920.2.camel@haakon3.risingtidesystems.com (mailing list archive)
State New, archived
Headers show

Commit Message

Nicholas A. Bellinger May 11, 2017, 5:21 p.m. UTC
On Thu, 2017-05-11 at 15:18 +0000, Bart Van Assche wrote:
> On Wed, 2017-05-10 at 23:46 -0700, Nicholas A. Bellinger wrote:
> > then I'll go ahead and push the revert here:
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending.git/commit/?h=for-next&id=69be17042db730092d25691dd85868a41d4d99ac
> 
> Hello Nic,
> 
> As you know any patch that is sent to Linus should be posted on the
> appropriate mailing list first. Posting an URL to a patch is not sufficient.
> 

It's a revert with more than appropriate context that has been discussed
ad-nauseum for the last number of days with no actual end-user impact.

But if you insist, here you go.

This reverts commit 0e2eb7d12eaa8e391bf5615d4271bb87a649caaa

  Author: Bart Van Assche <bart.vanassche@sandisk.com>
  Date:   Thu Mar 30 10:12:39 2017 -0700

      target: Fix VERIFY and WRITE VERIFY command parsing

This patch broke existing behaviour for WRITE_VERIFY because
it dropped the original SCF_SCSI_DATA_CDB assignment for
bytchk = 0 so target_cmd_size_check() no longer rejected
this case, allowing an overflow case to trigger an OOPs
in iscsi-target.

Since the short term and long term fixes are still being
discussed, revert it for now since it's late in the merge
window and try again in v4.13-rc1.

Conflicts:
	drivers/target/target_core_sbc.c

Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
---
 drivers/target/target_core_sbc.c | 74 ++++++----------------------------------
 1 file changed, 10 insertions(+), 64 deletions(-)


--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

Bart Van Assche May 11, 2017, 6:39 p.m. UTC | #1
On Thu, 2017-05-11 at 10:21 -0700, Nicholas A. Bellinger wrote:
> This reverts commit 0e2eb7d12eaa8e391bf5615d4271bb87a649caaa
> 
>   Author: Bart Van Assche <bart.vanassche@sandisk.com>
>   Date:   Thu Mar 30 10:12:39 2017 -0700
> 
>       target: Fix VERIFY and WRITE VERIFY command parsing
> 
> This patch broke existing behaviour for WRITE_VERIFY because
> it dropped the original SCF_SCSI_DATA_CDB assignment for
> bytchk = 0 so target_cmd_size_check() no longer rejected
> this case, allowing an overflow case to trigger an OOPs
> in iscsi-target.
> 
> Since the short term and long term fixes are still being
> discussed, revert it for now since it's late in the merge
> window and try again in v4.13-rc1.

Hello Nic,

In transport_generic_new_cmd(), called from iscsit_process_scsi_cmd() to
submit an iSCSI command to the LIO core, one can see that target_alloc_sgls()
is called whether or not SCF_SCSI_DATA_CDB has been set. So I don't think
that what you wrote in the patch description is correct. If you have a look
at my v2 patch series you will see that the buffer overflow I reported can
be triggered for any SCSI command and not just for VERIFY / WRITE AND VERIFY.
All that's needed to trigger a buffer overflow with the iSCSI target driver
is to send about 4 * PAGE_SIZE bytes more immediate data than the Data-Out
buffer size specified through the CDB.

Bart.--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Nicholas A. Bellinger May 11, 2017, 6:52 p.m. UTC | #2
On Thu, 2017-05-11 at 18:39 +0000, Bart Van Assche wrote:
> On Thu, 2017-05-11 at 10:21 -0700, Nicholas A. Bellinger wrote:
> > This reverts commit 0e2eb7d12eaa8e391bf5615d4271bb87a649caaa
> > 
> >   Author: Bart Van Assche <bart.vanassche@sandisk.com>
> >   Date:   Thu Mar 30 10:12:39 2017 -0700
> > 
> >       target: Fix VERIFY and WRITE VERIFY command parsing
> > 
> > This patch broke existing behaviour for WRITE_VERIFY because
> > it dropped the original SCF_SCSI_DATA_CDB assignment for
> > bytchk = 0 so target_cmd_size_check() no longer rejected
> > this case, allowing an overflow case to trigger an OOPs
> > in iscsi-target.
> > 
> > Since the short term and long term fixes are still being
> > discussed, revert it for now since it's late in the merge
> > window and try again in v4.13-rc1.
> 
> Hello Nic,
> 
> In transport_generic_new_cmd(), called from iscsit_process_scsi_cmd() to
> submit an iSCSI command to the LIO core, one can see that target_alloc_sgls()
> is called whether or not SCF_SCSI_DATA_CDB has been set.

Wrong.  iscsit_process_scsi_cmd() returns '1' to signal dump_payload to
iscsit_get_immediate_data() when any CDB with SCF_SCSI_DATA_CDB is set.

>  So I don't think
> that what you wrote in the patch description is correct. If you have a look
> at my v2 patch series you will see that the buffer overflow I reported can
> be triggered for any SCSI command and not just for VERIFY / WRITE AND VERIFY.
> All that's needed to trigger a buffer overflow with the iSCSI target driver
> is to send about 4 * PAGE_SIZE bytes more immediate data than the Data-Out
> buffer size specified through the CDB.

Considering a full page is still allocated for non SCF_SCSI_DATA_CDB,
and iscsit_map_iovec() still checks for SGL array overflow, I don't
think you'll be able to trigger anything.

If you can, you should post the test case to prove it.  :)

Regardless, I'm going to push the change post -rc1 to address the
regression that Roland's v4.3.y change introduced wrt to control WRITE
CDB overflow:

http://marc.info/?l=linux-scsi&m=149452576930990&w=2

Like I said earlier, you are more than welcome to add full WRITE CDB
overflow support, but since it's a new feature it's not v4.12 material.

--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/target/target_core_sbc.c b/drivers/target/target_core_sbc.c
index a0ad618..4316f7b 100644
--- a/drivers/target/target_core_sbc.c
+++ b/drivers/target/target_core_sbc.c
@@ -831,60 +831,6 @@  static sense_reason_t compare_and_write_callback(struct se_cmd *cmd, bool succes
 	return 0;
 }
 
-/**
- * sbc_parse_verify - parse VERIFY, VERIFY_16 and WRITE VERIFY commands
- * @cmd:     (in)  structure that describes the SCSI command to be parsed.
- * @sectors: (out) Number of logical blocks on the storage medium that will be
- *           affected by the SCSI command.
- * @bufflen: (out) Expected length of the SCSI Data-Out buffer.
- */
-static sense_reason_t sbc_parse_verify(struct se_cmd *cmd, int *sectors,
-				       u32 *bufflen)
-{
-	struct se_device *dev = cmd->se_dev;
-	u8 *cdb = cmd->t_task_cdb;
-	u8 bytchk = (cdb[1] >> 1) & 3;
-	sense_reason_t ret;
-
-	switch (cdb[0]) {
-	case VERIFY:
-	case WRITE_VERIFY:
-		*sectors = transport_get_sectors_10(cdb);
-		cmd->t_task_lba = transport_lba_32(cdb);
-		break;
-	case VERIFY_16:
-	case WRITE_VERIFY_16:
-		*sectors = transport_get_sectors_16(cdb);
-		cmd->t_task_lba = transport_lba_64(cdb);
-		break;
-	default:
-		WARN_ON_ONCE(true);
-		return TCM_UNSUPPORTED_SCSI_OPCODE;
-	}
-
-	if (sbc_check_dpofua(dev, cmd, cdb))
-		return TCM_INVALID_CDB_FIELD;
-
-	ret = sbc_check_prot(dev, cmd, cdb, *sectors, true);
-	if (ret)
-		return ret;
-
-	switch (bytchk) {
-	case 0:
-		*bufflen = 0;
-		break;
-	case 1:
-		*bufflen = sbc_get_size(cmd, *sectors);
-		cmd->se_cmd_flags |= SCF_SCSI_DATA_CDB;
-		break;
-	default:
-		pr_err("Unsupported BYTCHK value %d for SCSI opcode %#x\n",
-		       bytchk, cdb[0]);
-		return TCM_INVALID_CDB_FIELD;
-	}
-	return TCM_NO_SENSE;
-}
-
 sense_reason_t
 sbc_parse_cdb(struct se_cmd *cmd, struct sbc_ops *ops)
 {
@@ -952,6 +898,7 @@  static sense_reason_t sbc_parse_verify(struct se_cmd *cmd, int *sectors,
 		cmd->execute_cmd = sbc_execute_rw;
 		break;
 	case WRITE_10:
+	case WRITE_VERIFY:
 		sectors = transport_get_sectors_10(cdb);
 		cmd->t_task_lba = transport_lba_32(cdb);
 
@@ -965,13 +912,6 @@  static sense_reason_t sbc_parse_verify(struct se_cmd *cmd, int *sectors,
 		cmd->se_cmd_flags |= SCF_SCSI_DATA_CDB;
 		cmd->execute_cmd = sbc_execute_rw;
 		break;
-	case WRITE_VERIFY:
-	case WRITE_VERIFY_16:
-		ret = sbc_parse_verify(cmd, &sectors, &size);
-		if (ret)
-			return ret;
-		cmd->execute_cmd = sbc_execute_rw;
-		goto check_lba;
 	case WRITE_12:
 		sectors = transport_get_sectors_12(cdb);
 		cmd->t_task_lba = transport_lba_32(cdb);
@@ -987,6 +927,7 @@  static sense_reason_t sbc_parse_verify(struct se_cmd *cmd, int *sectors,
 		cmd->execute_cmd = sbc_execute_rw;
 		break;
 	case WRITE_16:
+	case WRITE_VERIFY_16:
 		sectors = transport_get_sectors_16(cdb);
 		cmd->t_task_lba = transport_lba_64(cdb);
 
@@ -1169,9 +1110,14 @@  static sense_reason_t sbc_parse_verify(struct se_cmd *cmd, int *sectors,
 		break;
 	case VERIFY:
 	case VERIFY_16:
-		ret = sbc_parse_verify(cmd, &sectors, &size);
-		if (ret)
-			return ret;
+		size = 0;
+		if (cdb[0] == VERIFY) {
+			sectors = transport_get_sectors_10(cdb);
+			cmd->t_task_lba = transport_lba_32(cdb);
+		} else {
+			sectors = transport_get_sectors_16(cdb);
+			cmd->t_task_lba = transport_lba_64(cdb);
+		}
 		cmd->execute_cmd = sbc_emulate_noop;
 		goto check_lba;
 	case REZERO_UNIT: