Skip to contentSkip to conversation

target: pscsi: avoid OOM in pscsi_map_sg()

PATCH Applied [email protected] 4 participants 7 messages Mar 23, 2021 – Mar 25, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/2 target: pscsi: avoid OOM in pscsi_map_sg()

From: Martin Wilck <[email protected]>

pscsi_map_sg() uses the variable nr_pages as a hint for bio_kmalloc() how many vector elements to allocate. If nr_pages is < BIO_MAX_PAGES, it will be reset to 0 after successful allocation of the bio.

If bio_add_pc_page() fails later for whatever reason, pscsi_map_sg() tries to allocate another bio, passing nr_vecs=0. This causes bio_add_pc_page() to fail immediately in the next call. pci_map_sg() continues to allocate zero-length bios until memory is exhausted and the kernel crashes with OOM. This can be easily observed by exporting a SATA DVD drive via pscsi. The target crashes as soon as the client tries to access the DVD LUN. In the case I analyzed, bio_add_pc_page() would fail because the DVD device's max_sectors_kb (128) was exceeded.

Avoid this by simply not resetting nr_pages to 0 after allocating the bio. This way, the client receives an IO error when it tries to send requests exceeding the devices max_sectors_kb, and eventually gets it right. The client must still limit max_sectors_kb e.g. by an udev rule if (like in my case) the driver doesn't report valid block limits, otherwise it encounters I/O errors.

---
 drivers/target/target_core_pscsi.c | 1 -
 1 file changed, 1 deletion(-)
drivers/target/target_core_pscsi.c +0 -1
index 7b1035e..977362d 100644
@@ -881,7 +881,6 @@ pscsi_map_sg(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
 			if (!bio) {
 new_bio:
 				nr_vecs = bio_max_segs(nr_pages);
-				nr_pages -= nr_vecs;
 				/*
 				 * Calls bio_kmalloc() and sets bio->bi_end_io()
 				 */
Signature · 1 line
2.26.2
#1 lore raw

PATCH v2 2/2 target: pscsi: cleanup after failure in pscsi_map_sg()

From: Martin Wilck <[email protected]>

If pscsi_map_sg() fails, make sure to drop references to already allocated bios.

---
 drivers/target/target_core_pscsi.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/target/target_core_pscsi.c +8 -0
index 977362d..1c9aeab 100644
@@ -937,6 +937,14 @@ pscsi_map_sg(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
 
 	return 0;
 fail:
+	if (bio)
+		bio_put(bio);
+	while (req->bio) {
+		bio = req->bio;
+		req->bio = bio->bi_next;
+		bio_put(bio);
+	}
+	req->biotail = NULL;
 	return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 }
 
Signature · 1 line
2.26.2
Reviewed-by on v2 2/2
#3 lore raw

Re: [PATCH v2 2/2] target: pscsi: cleanup after failure in pscsi_map_sg()

In reply to [email protected]
Quoted 20 lines · On 3/23/21 2:24 PM, [email protected] wrote:

From: Martin Wilck <[email protected]>

If pscsi_map_sg() fails, make sure to drop references to already allocated bios.

---
 drivers/target/target_core_pscsi.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/target/target_core_pscsi.c +1 -0
index 977362d..1c9aeab 100644
@@ -937,6 +937,14 @@ pscsi_map_sg(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
 
 	return 0;
 fail:
+	if (bio)

drivers/target/target_core_pscsi.c pscsi_map_sg(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents, · line 950

Quoted 10 lines
drivers/target/target_core_pscsi.c +7 -0
+		bio_put(bio);
+	while (req->bio) {
+		bio = req->bio;
+		req->bio = bio->bi_next;
+		bio_put(bio);
+	}
+	req->biotail = NULL;
 	return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 }
 
Reviewed-by on v2 1/2
#4 lore raw
In reply to [email protected]
Quoted 16 lines · On Tue, Mar 23, 2021 at 10:24:30PM +0100, [email protected] wrote:

From: Martin Wilck <[email protected]>

pscsi_map_sg() uses the variable nr_pages as a hint for bio_kmalloc() how many vector elements to allocate. If nr_pages is < BIO_MAX_PAGES, it will be reset to 0 after successful allocation of the bio.

If bio_add_pc_page() fails later for whatever reason, pscsi_map_sg() tries to allocate another bio, passing nr_vecs=0. This causes bio_add_pc_page() to fail immediately in the next call. pci_map_sg() continues to allocate zero-length bios until memory is exhausted and the kernel crashes with OOM. This can be easily observed by exporting a SATA DVD drive via pscsi. The target crashes as soon as the client tries to access the DVD LUN. In the case I analyzed, bio_add_pc_page() would fail because the DVD device's max_sectors_kb (128) was exceeded.

Quoted 8 lines

Avoid this by simply not resetting nr_pages to 0 after allocating the bio. This way, the client receives an IO error when it tries to send requests exceeding the devices max_sectors_kb, and eventually gets it right. The client must still limit max_sectors_kb e.g. by an udev rule if (like in my case) the driver doesn't report valid block limits, otherwise it encounters I/O errors.

Looks good,

Reviewed-by on v2 1/2
#5 lore raw
In reply to [email protected]
Quoted 34 lines · On 3/23/21 2:24 PM, [email protected] wrote:

From: Martin Wilck <[email protected]>

pscsi_map_sg() uses the variable nr_pages as a hint for bio_kmalloc() how many vector elements to allocate. If nr_pages is < BIO_MAX_PAGES, it will be reset to 0 after successful allocation of the bio.

If bio_add_pc_page() fails later for whatever reason, pscsi_map_sg() tries to allocate another bio, passing nr_vecs=0. This causes bio_add_pc_page() to fail immediately in the next call. pci_map_sg() continues to allocate zero-length bios until memory is exhausted and the kernel crashes with OOM. This can be easily observed by exporting a SATA DVD drive via pscsi. The target crashes as soon as the client tries to access the DVD LUN. In the case I analyzed, bio_add_pc_page() would fail because the DVD device's max_sectors_kb (128) was exceeded.

Avoid this by simply not resetting nr_pages to 0 after allocating the bio. This way, the client receives an IO error when it tries to send requests exceeding the devices max_sectors_kb, and eventually gets it right. The client must still limit max_sectors_kb e.g. by an udev rule if (like in my case) the driver doesn't report valid block limits, otherwise it encounters I/O errors.

---
 drivers/target/target_core_pscsi.c | 1 -
 1 file changed, 1 deletion(-)
drivers/target/target_core_pscsi.c +0 -0
index 7b1035e..977362d 100644

drivers/target/target_core_pscsi.c pscsi_map_sg(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents, · line 886

Quoted 7 lines
drivers/target/target_core_pscsi.c +0 -1
@@ -881,7 +881,6 @@ pscsi_map_sg(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
 			if (!bio) {
 new_bio:
 				nr_vecs = bio_max_segs(nr_pages);
-				nr_pages -= nr_vecs;
 				/*
 				 * Calls bio_kmalloc() and sets bio->bi_end_io()
 				 */
#6 lore raw
In reply to [email protected]
Quoted 14 lines · On Tue, 23 Mar 2021 22:24:30 +0100, [email protected] wrote:

pscsi_map_sg() uses the variable nr_pages as a hint for bio_kmalloc() how many vector elements to allocate. If nr_pages is < BIO_MAX_PAGES, it will be reset to 0 after successful allocation of the bio.

If bio_add_pc_page() fails later for whatever reason, pscsi_map_sg() tries to allocate another bio, passing nr_vecs=0. This causes bio_add_pc_page() to fail immediately in the next call. pci_map_sg() continues to allocate zero-length bios until memory is exhausted and the kernel crashes with OOM. This can be easily observed by exporting a SATA DVD drive via pscsi. The target crashes as soon as the client tries to access the DVD LUN. In the case I analyzed, bio_add_pc_page() would fail because the DVD device's max_sectors_kb (128) was exceeded.

Quoted 1 line

[...]

Applied to 5.12/scsi-fixes, thanks!

[1/2] target: pscsi: avoid OOM in pscsi_map_sg()

      https://git.kernel.org/mkp/scsi/c/077ce028b8e0
[2/2] target: pscsi: cleanup after failure in pscsi_map_sg()
      https://git.kernel.org/mkp/scsi/c/36fa766faa0c
Signature · 1 line
Martin K. Petersen	Oracle Linux Engineering