lib/tst_cgroup: fix short reads of mems/cpus
PATCH 1/1 lib/tst_cgroup: fix short reads of mems/cpus
tst_cgroup_cpuset_read_files() is wrongly using sizeof on pointer type to figure out length of buffer. On some systems there's more content in cgroup mems/cpus, which leads to partial read and subsequent EINVAL on write.
Also the retval buffer should be zero terminated, since buffer can be passed uninitialized, subsequently leading to writing garbage to cgroup and again hitting EINVAL.
Before:
tst_test.c:1250: TINFO: Timeout per run is 0h 05m 00s
tst_device.c:423: TINFO: No device is mounted at cgroup2
tst_device.c:423: TINFO: No device is mounted at /tmp/cgroup_cst
tst_cgroup.c:100: TINFO: Cgroup(cpuset) v1 mount at /tmp/cgroup_cst success safe_macros.c:458: TBROK: tst_cgroup.c:449: write(6,0x7fffc0425410,18446744073709551615) failed: EINVAL (22) tst_cgroup.c:173: TWARN: umount /tmp/cgroup_cst: EBUSY (16)
tst_cgroup.c:175: TWARN: rmdir /tmp/cgroup_cst: EBUSY (16)
tst_cgroup.c:178: TINFO: Cgroup v1 unmount success
After:
tst_test.c:1250: TINFO: Timeout per run is 0h 05m 00s tst_device.c:423: TINFO: No device is mounted at cgroup2 tst_device.c:423: TINFO: No device is mounted at /tmp/cgroup_cst tst_cgroup.c:100: TINFO: Cgroup(cpuset) v1 mount at /tmp/cgroup_cst success cpuset01.c:83: TPASS: cpuset test pass
tst_cgroup.c:178: TINFO: Cgroup v1 unmount success
--- include/tst_cgroup.h | 6 ++++-- lib/tst_cgroup.c | 6 ++++-- testcases/kernel/mem/cpuset/cpuset01.c | 4 ++-- 3 files changed, 10 insertions(+), 6 deletions(-)
include/tst_cgroup.h +4 -2
@@ -39,7 +39,9 @@ int tst_cgroup_mem_swapacct_enabled(const char *cgroup_dir);void tst_cgroup_mem_set_maxswap (const char *cgroup_dir,long memsz);/* Set of functions to read/write cpuset controller files content */ -void tst_cgroup_cpuset_read_files (const char *cgroup_dir,const char *filename,char *retbuf); -void tst_cgroup_cpuset_write_files (const char *cgroup_dir,const char *filename,const char *buf); +void tst_cgroup_cpuset_read_files (const char *cgroup_dir,const char *filename, +char *retbuf,size_t retbuf_sz); +void tst_cgroup_cpuset_write_files (const char *cgroup_dir,const char *filename, +const char *buf);#endif /* TST_CGROUP_H */
lib/tst_cgroup.c +4 -2
@@ -393,7 +393,8 @@ void tst_cgroup_mem_set_maxswap(const char *cgroup_dir, long memsz)tst_cgroup_set_knob (cgroup_dir,"memory.swap.max" , memsz); } -void tst_cgroup_cpuset_read_files (const char *cgroup_dir,const char *filename,char *retbuf) +void tst_cgroup_cpuset_read_files (const char *cgroup_dir,const char *filename, +char *retbuf,size_t retbuf_sz) {int fd;char *cgroup_new_dir;
@@ -417,7 +418,8 @@ void tst_cgroup_cpuset_read_files(const char *cgroup_dir, const char *filename,tst_brk (TBROK |TERRNO ,"open %s" , knob_path); } -if (read (fd, retbuf,sizeof (retbuf)) <0 ) +memset (retbuf,0 , retbuf_sz); +if (read (fd, retbuf, retbuf_sz) <0 )tst_brk (TBROK |TERRNO ,"read %s" , knob_path);close (fd);
testcases/kernel/mem/cpuset/cpuset01.c +2 -2
@@ -51,9 +51,9 @@ static void test_cpuset(void)unsigned long nmask[MAXNODES /BITS_PER_LONG ] = {0 };char mems[BUFSIZ ], buf[BUFSIZ ]; -tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"cpus" , buf); +tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"cpus" , buf,sizeof (buf));tst_cgroup_cpuset_write_files (PATH_TMP_CG_CST ,"cpus" , buf); -tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"mems" , mems); +tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"mems" , mems,sizeof (buf));tst_cgroup_cpuset_write_files (PATH_TMP_CG_CST ,"mems" , mems);tst_cgroup_move_current (PATH_TMP_CG_CST );
Signature · 1 line
2.18.1
PATCH 1/1 lib/tst_cgroup: fix short reads of mems/cpus
Quoted 93 lines · Jan Stancek
tst_cgroup_cpuset_read_files()is wrongly using sizeof on pointer type to figure out length of buffer. On some systems there's more content in cgroup mems/cpus, which leads to partial read and subsequent EINVAL on write.Also the retval buffer should be zero terminated, since buffer can be passed uninitialized, subsequently leading to writing garbage to cgroup and again hitting EINVAL.
Before:
tst_test.c:1250: TINFO: Timeout per run is 0h 05m 00s tst_device.c:423: TINFO: No device is mounted at cgroup2 tst_device.c:423: TINFO: No device is mounted at/tmp/cgroup_csttst_cgroup.c:100: TINFO: Cgroup(cpuset) v1 mount at/tmp/cgroup_cst
success
safe_macros.c:458: TBROK: tst_cgroup.c:449:write(6,0x7fffc0425410,18446744073709551615)failed: EINVAL (22)
tst_cgroup.c:173: TWARN: umount/tmp/cgroup_cst: EBUSY (16) tst_cgroup.c:175: TWARN: rmdir/tmp/cgroup_cst: EBUSY (16) tst_cgroup.c:178: TINFO: Cgroup v1 unmount successAfter:
tst_test.c:1250: TINFO: Timeout per run is 0h 05m 00s tst_device.c:423: TINFO: No device is mounted at cgroup2 tst_device.c:423: TINFO: No device is mounted at/tmp/cgroup_csttst_cgroup.c:100: TINFO: Cgroup(cpuset) v1 mount at/tmp/cgroup_cst
success
cpuset01.c:83: TPASS: cpuset test pass
tst_cgroup.c:178: TINFO: Cgroup v1 unmount successSigned-off-by Jan Stancek <[email protected]>--- include/tst_cgroup.h | 6 ++++-- lib/tst_cgroup.c | 6 ++++-- testcases/kernel/mem/cpuset/cpuset01.c | 4 ++-- 3 files changed, 10 insertions(+), 6 deletions(-)include/tst_cgroup.h +0
-2@@ -39,7 +39,9 @@ int tst_cgroup_mem_swapacct_enabled(const char *cgroup_dir);void tst_cgroup_mem_set_maxswap (const char *cgroup_dir,long memsz);/* Set of functions to read/write cpuset controller files content */ -void tst_cgroup_cpuset_read_files (const char *cgroup_dir,const char *filename, char *retbuf); -void tst_cgroup_cpuset_write_files(const char *cgroup_dir, const char*filename, const char *buf);
+void tst_cgroup_cpuset_read_files(const char*cgroup_dir, const char *filename,include/tst_cgroup.h +2
-0+ char *retbuf, size_t retbuf_sz); +void tst_cgroup_cpuset_write_files(const char *cgroup_dir, const char*filename,
include/tst_cgroup.h +1
-0+ const char *buf);#endif /* TST_CGROUP_H */ lib/tst_cgroup.c +2
-1@@ -393,7 +393,8 @@ void tst_cgroup_mem_set_maxswap(const char *cgroup_dir, long memsz)tst_cgroup_set_knob (cgroup_dir,"memory.swap.max" , memsz); } -void tst_cgroup_cpuset_read_files (const char *cgroup_dir,const char *filename, char *retbuf) +void tst_cgroup_cpuset_read_files (const char *cgroup_dir,const char *filename, +char *retbuf,size_t retbuf_sz){ int fd; char *cgroup_new_dir;lib/tst_cgroup.c +2
-1@@ -417,7 +418,8 @@ void tst_cgroup_cpuset_read_files(const char *cgroup_dir, const char *filename,tst_brk (TBROK |TERRNO ,"open %s" , knob_path); } -if (read (fd, retbuf,sizeof (retbuf)) <0 ) +memset (retbuf,0 , retbuf_sz); +if (read (fd, retbuf, retbuf_sz) <0 )tst_brk (TBROK |TERRNO ,"read %s" , knob_path);close(fd);testcases/kernel/mem/cpuset/cpuset01.c +0
-0@@ -51,9 +51,9 @@ static void test_cpuset(void)unsigned long nmask[MAXNODES /BITS_PER_LONG ] = {0 };
testcases/kernel/mem/cpuset/cpuset01.c static void test_cpuset(void) · line 59
Quoted 10 lines
testcases/kernel/mem/cpuset/cpuset01.c +2
-2char mems[BUFSIZ ], buf[BUFSIZ ]; -tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"cpus" , buf); +tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"cpus" , buf,sizeof (buf));tst_cgroup_cpuset_write_files (PATH_TMP_CG_CST ,"cpus" , buf); -tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"mems" , mems); +tst_cgroup_cpuset_read_files (PATH_TMP_CG_CST ,"mems" , mems,sizeof (buf));
sizeof() is generally used to calculate the size (in bytes) of a data type, e.g sizeof(char *).
I think here to pass 'BUFSIZ' directly is better than sizeof(buf).
Otherwise, looks good to me.
Quoted 7 lines
tst_cgroup_cpuset_write_files(PATH_TMP_CG_CST, "mems", mems); tst_cgroup_move_current(PATH_TMP_CG_CST);--
2.18.1
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20201104/ea1346c6/attachment.htm>
PATCH 1/1 lib/tst_cgroup: fix short reads of mems/cpus
Quoted 10 lines · Li Wang
...
Quoted 8 lines
(quoted diff) +1
-1- tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "cpus", buf); + tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "cpus", buf,sizeof(buf)); tst_cgroup_cpuset_write_files(PATH_TMP_CG_CST, "cpus", buf);(quoted diff) +1
-1- tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "mems", mems); + tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "mems", mems,sizeof(buf));
Quoted 4 lines
sizeof()is generally used to calculate the size (in bytes) of a data type, e.gsizeof(char *).
I think here to pass 'BUFSIZ' directly is better than sizeof(buf).
Jan, do you agree with this point?
If yes, I can help to modify and apply this patch.
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20201109/b2c2e584/attachment.htm>
PATCH 1/1 lib/tst_cgroup: fix short reads of mems/cpus
----- Original Message -----
Quoted 12 lines · Li Wang
Quoted 11 lines
...
Quoted 8 lines
(quoted diff) +1
-1- tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "cpus", buf); + tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "cpus", buf,sizeof(buf)); tst_cgroup_cpuset_write_files(PATH_TMP_CG_CST, "cpus", buf);(quoted diff) +1
-1- tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "mems", mems); + tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "mems", mems,sizeof(buf));
Quoted 7 lines
Quoted 4 lines
sizeof()is generally used to calculate the size (in bytes) of a data type, e.gsizeof(char *).
I think here to pass 'BUFSIZ' directly is better than sizeof(buf).Jan, do you agree with this point?
If yes, I can help to modify and apply this patch.
2nd line should have been sizeof(mems), but BUFSIZ will work too.
PATCH 1/1 lib/tst_cgroup: fix short reads of mems/cpus
Quoted 17 lines · Jan Stancek
----- Original Message -----
Quoted 12 lines · Li Wang
Quoted 11 lines
...
Quoted 8 lines
(quoted diff) +1
-1- tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "cpus", buf); + tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "cpus", buf,sizeof(buf)); tst_cgroup_cpuset_write_files(PATH_TMP_CG_CST, "cpus", buf);(quoted diff) +1
-1- tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "mems", mems); + tst_cgroup_cpuset_read_files(PATH_TMP_CG_CST, "mems", mems,sizeof(buf));
Quoted 1 line
Quoted 1 line
sizeof()is generally used to calculate the size (in bytes) of a datatype,
Quoted 9 lines
Quoted 6 lines
Quoted 3 lines
e.g
sizeof(char *).
I think here to pass 'BUFSIZ' directly is better than sizeof(buf).Jan, do you agree with this point?
If yes, I can help to modify and apply this patch.2nd line should have been sizeof(mems), but BUFSIZ will work too.
Pushed, thanks!
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20201110/0ea7c10f/attachment-0001.htm>