Skip to contentSkip to conversation

perf test: Fix test trace+probe_vfs_getname.sh on s390

PATCH Applied Thomas Richter 4 participants 4 messages Feb 17, 2020 – Feb 26, 2020 perf events

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 perf test: Fix test trace+probe_vfs_getname.sh on s390

This test places a kprobe to function getname_flags() in the kernel which has the following prototype:

struct filename *
getname_flags(const char __user *filename, int flags, int *empty)

Variable filename points to a filename located in user space memory. Looking at
commit 88903c464321c ("tracing/probe: Add ustring type for user-space string") the kprobe should indicate that user space memory is accessed.

Output before:
[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : FAILED!
67: Check open filename arg using perf trace + vfs_getname: FAILED! [root@m35lp76 perf]#

Output after:
[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : Ok
67: Check open filename arg using perf trace + vfs_getname: Ok [root@m35lp76 perf]#

Comments from Masami Hiramatsu:
This bug doesn't happen on x86 or other archs on which user-address space and kernel address space is same. On some arch (ppc64 in this case?) user-address space is partially or completely same as kernel address space. (Yes, they switch the world when running into the kernel) In this case, we need to use different data access functions for each spaces. That is why I introduced "ustring" type for kprobe event.
As far as I can see, Thomas's patch is sane. Thomas, could you show us your result on your test environment?
Thank you

Comments from Thomas Richter:
Test results included above.

---
 tools/perf/tests/shell/lib/probe_vfs_getname.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/tests/shell/lib/probe_vfs_getname.sh +1 -1
index 7cb99b433888..c2cc42daf924 100644
@@ -14,7 +14,7 @@ add_probe_vfs_getname() {
 	if [ $had_vfs_getname -eq 1 ] ; then
 		line=$(perf probe -L getname_flags 2>&1 | egrep 'result.*=.*filename;' | sed -r 's/[[:space:]]+([[:digit:]]+)[[:space:]]+result->uptr.*/\1/')
 		perf probe -q       "vfs_getname=getname_flags:${line} pathname=result->name:string" || \
-		perf probe $verbose "vfs_getname=getname_flags:${line} pathname=filename:string"
+		perf probe $verbose "vfs_getname=getname_flags:${line} pathname=filename:ustring"
 	fi
 }
 
Signature · 1 line
2.21.0
Acked-by on v4 1/1
#1 lore raw
Quoted 34 lines · On Mon, 17 Feb 2020 11:21:11 +0100 Thomas Richter <[email protected]> wrote:

This test places a kprobe to function getname_flags() in the kernel which has the following prototype:

struct filename *
getname_flags(const char __user *filename, int flags, int *empty)

Variable filename points to a filename located in user space memory. Looking at
commit 88903c464321c ("tracing/probe: Add ustring type for user-space string") the kprobe should indicate that user space memory is accessed.

Output before:
[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : FAILED!
67: Check open filename arg using perf trace + vfs_getname: FAILED! [root@m35lp76 perf]#

Output after:
[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : Ok
67: Check open filename arg using perf trace + vfs_getname: Ok [root@m35lp76 perf]#

Comments from Masami Hiramatsu:
This bug doesn't happen on x86 or other archs on which user-address space and kernel address space is same. On some arch (ppc64 in this case?) user-address space is partially or completely same as kernel address space. (Yes, they switch the world when running into the kernel) In this case, we need to use different data access functions for each spaces. That is why I introduced "ustring" type for kprobe event.
As far as I can see, Thomas's patch is sane. Thomas, could you show us your result on your test environment?
Thank you

Quoted 4 lines

Comments from Thomas Richter:
Test results included above.

Looks good to me.

Thank you,

Quoted 20 lines
---
 tools/perf/tests/shell/lib/probe_vfs_getname.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/tests/shell/lib/probe_vfs_getname.sh +1 -1
index 7cb99b433888..c2cc42daf924 100644
@@ -14,7 +14,7 @@ add_probe_vfs_getname() {
 	if [ $had_vfs_getname -eq 1 ] ; then
 		line=$(perf probe -L getname_flags 2>&1 | egrep 'result.*=.*filename;' | sed -r 's/[[:space:]]+([[:digit:]]+)[[:space:]]+result->uptr.*/\1/')
 		perf probe -q       "vfs_getname=getname_flags:${line} pathname=result->name:string" || \
-		perf probe $verbose "vfs_getname=getname_flags:${line} pathname=filename:string"
+		perf probe $verbose "vfs_getname=getname_flags:${line} pathname=filename:ustring"
 	fi
 }
 
Signature · 1 line
2.21.0
Signature · 1 line
Masami Hiramatsu <[email protected]>
#2 lore raw

Em Mon, Feb 17, 2020 at 09:10:10PM +0900, Masami Hiramatsu escreveu:

Quoted 37 lines
Quoted 34 lines · On Mon, 17 Feb 2020 11:21:11 +0100 Thomas Richter <[email protected]> wrote:

This test places a kprobe to function getname_flags() in the kernel which has the following prototype:

struct filename *
getname_flags(const char __user *filename, int flags, int *empty)

Variable filename points to a filename located in user space memory. Looking at
commit 88903c464321c ("tracing/probe: Add ustring type for user-space string") the kprobe should indicate that user space memory is accessed.

Output before:
[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : FAILED!
67: Check open filename arg using perf trace + vfs_getname: FAILED! [root@m35lp76 perf]#

Output after:
[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : Ok
67: Check open filename arg using perf trace + vfs_getname: Ok [root@m35lp76 perf]#

Comments from Masami Hiramatsu:
This bug doesn't happen on x86 or other archs on which user-address space and kernel address space is same. On some arch (ppc64 in this case?) user-address space is partially or completely same as kernel address space. (Yes, they switch the world when running into the kernel) In this case, we need to use different data access functions for each spaces. That is why I introduced "ustring" type for kprobe event.
As far as I can see, Thomas's patch is sane. Thomas, could you show us your result on your test environment?
Thank you

Quoted 8 lines
Quoted 4 lines

Comments from Thomas Richter:
Test results included above.

Looks good to me.

Thanks, I tested it successfully on x86-64, applying.

- Arnaldo

merge bot Applied
#3 lore raw
In reply to Thomas Richter
The following commit has been merged into the perf/urgent branch of tip:

The following commit has been merged into the perf/urgent branch of tip:

Commit-ID:     2bbc83537614517730e9f2811195004b712de207
Gitweb:        https://git.kernel.org/tip/2bbc83537614517730e9f2811195004b712de207
Author:        Thomas Richter <[email protected]>
AuthorDate:    Mon, 17 Feb 2020 11:21:11 +01:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Tue, 18 Feb 2020 10:13:28 -03:00

perf test: Fix test trace+probe_vfs_getname.sh on s390

This test places a kprobe to function getname_flags() in the kernel which has the following prototype:

struct filename *getname_flags(const char __user *filename, int flags, int *empty)

The 'filename' argument points to a filename located in user space memory.

Looking at commit 88903c464321c ("tracing/probe: Add ustring type for user-space string") the kprobe should indicate that user space memory is accessed.

Output before:

[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : FAILED!
67: Check open filename arg using perf trace + vfs_getname: FAILED! [root@m35lp76 perf]#

Output after:

[root@m35lp76 perf]# ./perf test 66 67
66: Use vfs_getname probe to get syscall args filenames : Ok
67: Check open filename arg using perf trace + vfs_getname: Ok [root@m35lp76 perf]#

Comments from Masami Hiramatsu:

This bug doesn't happen on x86 or other archs on which user address space and kernel address space is the same. On some arches (ppc64 in this case?) user address space is partially or completely the same as kernel address space.

(Yes, they switch the world when running into the kernel) In this case, we need to use different data access functions for each space.

That is why I introduced the "ustring" type for kprobe events.

As far as I can see, Thomas's patch is sane. Thomas, could you show us your result on your test environment?

Comments from Thomas Richter:

Test results for s/390 included above.

---
 tools/perf/tests/shell/lib/probe_vfs_getname.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/tests/shell/lib/probe_vfs_getname.sh +1 -1
index 7cb99b4..c2cc42d 100644
@@ -14,7 +14,7 @@ add_probe_vfs_getname() {
 	if [ $had_vfs_getname -eq 1 ] ; then
 		line=$(perf probe -L getname_flags 2>&1 | egrep 'result.*=.*filename;' | sed -r 's/[[:space:]]+([[:digit:]]+)[[:space:]]+result->uptr.*/\1/')
 		perf probe -q       "vfs_getname=getname_flags:${line} pathname=result->name:string" || \
-		perf probe $verbose "vfs_getname=getname_flags:${line} pathname=filename:string"
+		perf probe $verbose "vfs_getname=getname_flags:${line} pathname=filename:ustring"
 	fi
 }