Skip to contentSkip to conversation

dasd_fba: Display '00000000' for zero page when dumping sense

PATCH In review Hannes Reinecke 2 participants 2 messages Nov 18, 2019 – Nov 21, 2019

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

PATCH 1/1 dasd_fba: Display '00000000' for zero page when dumping sense

When a discard I/O fails, dasd_fba_dump_sense() will crash as it tries to print out the CCW, and failing to take into account that for discard I/O we have only one data pointer, not one per sg. As the data pointer will always point to the zero page this patch replaces the data pointer output with '00000000' to avoid the crash.

[sparschauer: replaced "ccw" with "act", "snprintf" with "sprintf"]
[sparschauer v2: added missing curly braces to for loops]
---
 drivers/s390/block/dasd_fba.c | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)
drivers/s390/block/dasd_fba.c +18 -3
index cbb770824226..4b867bd6b164 100644
@@ -717,10 +717,15 @@ dasd_fba_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req,
 			       " CCW %p: %08X %08X DAT:",
 			       act, ((int *) act)[0], ((int *) act)[1]);
 		for (count = 0; count < 32 && count < act->count;
-		     count += sizeof(int))
+		     count += sizeof(int)) {
+			if (act->flags & CCW_FLAG_SLI) {
+				len += sprintf(page + len, " 00000000");
+				break;
+			}
 			len += sprintf(page + len, " %08X",
 				       ((int *) (addr_t) act->cda)
 				       [(count>>2)]);
+		}
 		len += sprintf(page + len, "\n");
 		act++;
 	}
@@ -739,10 +744,15 @@ dasd_fba_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req,
 			       " CCW %p: %08X %08X DAT:",
 			       act, ((int *) act)[0], ((int *) act)[1]);
 		for (count = 0; count < 32 && count < act->count;
-		     count += sizeof(int))
+		     count += sizeof(int)) {
+			if (act->flags & CCW_FLAG_SLI) {
+				len += sprintf(page + len, " 00000000");
+				break;
+			}
 			len += sprintf(page + len, " %08X",
 				       ((int *) (addr_t) act->cda)
 				       [(count>>2)]);
+		}
 		len += sprintf(page + len, "\n");
 		act++;
 	}
@@ -757,10 +767,15 @@ dasd_fba_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req,
 			       " CCW %p: %08X %08X DAT:",
 			       act, ((int *) act)[0], ((int *) act)[1]);
 		for (count = 0; count < 32 && count < act->count;
-		     count += sizeof(int))
+		     count += sizeof(int)) {
+			if (act->flags & CCW_FLAG_SLI) {
+				len += sprintf(page + len, " 00000000");
+				break;
+			}
 			len += sprintf(page + len, " %08X",
 				       ((int *) (addr_t) act->cda)
 				       [(count>>2)]);
+		}
 		len += sprintf(page + len, "\n");
 		act++;
 	}
Signature · 1 line
2.16.4
maintainer
#1 lore raw
Quoted 6 lines · Hannes Reinecke

When a discard I/O fails, dasd_fba_dump_sense() will crash as it tries to print out the CCW, and failing to take into account that for discard I/O we have only one data pointer, not one per sg. As the data pointer will always point to the zero page this patch replaces the data pointer output with '00000000' to avoid the crash.

Quoted 8 lines
[sparschauer: replaced "ccw" with "act", "snprintf" with "sprintf"]
[sparschauer v2: added missing curly braces to for loops]
---
 drivers/s390/block/dasd_fba.c | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)

Hi Hannes,

thanks for the patch. However,

drivers/s390/block/dasd_fba.c dasd_fba_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req, · line 721

Quoted 11 lines
drivers/s390/block/dasd_fba.c +2 -1
index cbb770824226..4b867bd6b164 100644
@@ -717,10 +717,15 @@ dasd_fba_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req,
 			       " CCW %p: %08X %08X DAT:",
 			       act, ((int *) act)[0], ((int *) act)[1]);
 		for (count = 0; count < 32 && count < act->count;
-		     count += sizeof(int))
+		     count += sizeof(int)) {
+			if (act->flags & CCW_FLAG_SLI) {

I'm not quite happy with the usage of CCW_FLAG_SLI here. We're currently looking into this issue in more detail to fully understand the problem.

I'll let you know the outcome and possible improvements for the patch as soon as possible.

regards,
Jan