Improve ptrace selftest usability
Subsystems (3)
cover Improve ptrace selftest usability
While trying to test changes to the breakpoint implementation in the kernel, I tried to run the ptrace tests and met many unexplained skips and failures.
This series addresses the pain points of trying to run these tests and learn what they are doing.
Benjamin Gray (4):
Documentation/powerpc: Fix ptrace request names selftests/powerpc/ptrace: Explain why tests are skipped selftests/powerpc/ptrace: Fix typo in pid_max search error selftests/powerpc/ptrace: Declare test temporary variables as volatile
Documentation/powerpc/ptrace.rst | 8 +++--- .../testing/selftests/powerpc/ptrace/child.h | 4 +-- .../selftests/powerpc/ptrace/core-pkey.c | 2 +- .../selftests/powerpc/ptrace/perf-hwbreak.c | 2 +- .../selftests/powerpc/ptrace/ptrace-hwbreak.c | 26 +++++++++---------- .../powerpc/ptrace/ptrace-perf-hwbreak.c | 6 ++--- .../selftests/powerpc/ptrace/ptrace-pkey.c | 2 +- .../selftests/powerpc/ptrace/ptrace-tar.c | 2 +- .../selftests/powerpc/ptrace/ptrace-tm-gpr.c | 4 +-- .../powerpc/ptrace/ptrace-tm-spd-gpr.c | 4 +-- .../powerpc/ptrace/ptrace-tm-spd-tar.c | 4 +-- .../powerpc/ptrace/ptrace-tm-spd-vsx.c | 4 +-- .../selftests/powerpc/ptrace/ptrace-tm-spr.c | 4 +-- .../selftests/powerpc/ptrace/ptrace-tm-tar.c | 4 +-- .../selftests/powerpc/ptrace/ptrace-tm-vsx.c | 4 +-- .../selftests/powerpc/ptrace/ptrace-vsx.c | 2 +- 16 files changed, 41 insertions(+), 41 deletions(-)
--
2.41.0
PATCH 1/4 Documentation/powerpc: Fix ptrace request names
The documented ptrace request names are currently wrong/incomplete. Fix this to improve correctness and searchability.
--- Documentation/powerpc/ptrace.rst | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
Documentation/powerpc/ptrace.rst +4 -4
@@ -15,7 +15,7 @@ that's extendable and that covers both BookE and server processors, so that GDB doesn't need to special-case each of them. We added the following 3 new ptrace requests. -1. PTRACE_PPC_GETHWDEBUGINFO +1. PPC_PTRACE_GETHWDBGINFO ============================ Query for GDB to discover the hardware debug features. The main info to
@@ -48,7 +48,7 @@ features will have bits indicating whether there is support for:: #define PPC_DEBUG_FEATURE_DATA_BP_DAWR 0x10 #define PPC_DEBUG_FEATURE_DATA_BP_ARCH_31 0x20 -2. PTRACE_SETHWDEBUG +2. PPC_PTRACE_SETHWDEBUG Sets a hardware breakpoint or watchpoint, according to the provided structure::
@@ -88,7 +88,7 @@ that the BookE supports. COMEFROM breakpoints available in server processors are not contemplated, but that is out of the scope of this work. ptrace will return an integer (handle) uniquely identifying the breakpoint or -watchpoint just created. This integer will be used in the PTRACE_DELHWDEBUG +watchpoint just created. This integer will be used in the PPC_PTRACE_DELHWDEBUG request to ask for its removal. Return -ENOSPC if the requested breakpoint can't be allocated on the registers.
@@ -150,7 +150,7 @@ Some examples of using the structure to: p.addr2 = (uint64_t) end_range; p.condition_value = 0; -3. PTRACE_DELHWDEBUG +3. PPC_PTRACE_DELHWDEBUG Takes an integer which identifies an existing breakpoint or watchpoint (i.e., the value returned from PTRACE_SETHWDEBUG), and deletes the
Signature · 1 line
2.41.0
PATCH 4/4 selftests/powerpc/ptrace: Declare test temporary variables as volatile
In reply to Benjamin GrayWhile the target is volatile, the temporary variables used to access the target cast away the volatile. This is undefined behaviour, and a compiler may optimise away/reorder these accesses, breaking the test.
This was observed with GCC 13.1.1, but it can be difficult to reproduce because of the dependency on compiler behaviour.
--- .../selftests/powerpc/ptrace/ptrace-hwbreak.c | 24 +++++++++---------- 1 file changed, 12 insertions(+), 12 deletions(-)
tools/testing/selftests/powerpc/ptrace/ptrace-hwbreak.c +12 -12
@@ -64,26 +64,26 @@ static bool dawr_present(struct ppc_debug_info *dbginfo)static void write_var (int len) { -__u8 *pcvar; -__u16 *psvar; -__u32 *pivar; -__u64 *plvar; +volatile __u8 *pcvar; +volatile __u16 *psvar; +volatile __u32 *pivar; +volatile __u64 *plvar;switch (len) {case 1 : - pcvar = (__u8 *)&glvar; + pcvar = (volatile __u8 *)&glvar; *pcvar =0xff ;break ;case 2 : - psvar = (__u16 *)&glvar; + psvar = (volatile __u16 *)&glvar; *psvar =0xffff ;break ;case 4 : - pivar = (__u32 *)&glvar; + pivar = (volatile __u32 *)&glvar; *pivar =0xffffffff ;break ;case 8 : - plvar = (__u64 *)&glvar; + plvar = (volatile __u64 *)&glvar; *plvar =0xffffffffffffffffLL ;break ; }
@@ -98,16 +98,16 @@ static void read_var(int len)switch (len) {case 1 : - cvar = (__u8 )glvar; + cvar = (volatile __u8 )glvar;break ;case 2 : - svar = (__u16 )glvar; + svar = (volatile __u16 )glvar;break ;case 4 : - ivar = (__u32 )glvar; + ivar = (volatile __u32 )glvar;break ;case 8 : - lvar = (__u64 )glvar; + lvar = (volatile __u64 )glvar;break ; } }
Signature · 1 line
2.41.0
PATCH 2/4 selftests/powerpc/ptrace: Explain why tests are skipped
In reply to Benjamin GrayMany tests require specific hardware features/configurations that a typical machine might not have. As a result, it's common to see a test is skipped. But it is tedious to find out why a test is skipped when all it gives is the file location of the skip macro.
Convert SKIP_IF() to SKIP_IF_MSG(), with appropriate descriptions of why the test is being skipped. This gives a general idea of why a test is skipped, which can be looked into further if it doesn't make sense.
--- tools/testing/selftests/powerpc/ptrace/child.h | 4 ++-- tools/testing/selftests/powerpc/ptrace/core-pkey.c | 2 +- tools/testing/selftests/powerpc/ptrace/perf-hwbreak.c | 2 +- tools/testing/selftests/powerpc/ptrace/ptrace-hwbreak.c | 2 +- tools/testing/selftests/powerpc/ptrace/ptrace-perf-hwbreak.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-pkey.c | 2 +- tools/testing/selftests/powerpc/ptrace/ptrace-tar.c | 2 +- tools/testing/selftests/powerpc/ptrace/ptrace-tm-gpr.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-tm-spd-gpr.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-tm-spd-tar.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-tm-spd-vsx.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-tm-spr.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-tm-tar.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-tm-vsx.c | 4 ++-- tools/testing/selftests/powerpc/ptrace/ptrace-vsx.c | 2 +- 15 files changed, 24 insertions(+), 24 deletions(-)
tools/testing/selftests/powerpc/ptrace/child.h +2 -2
@@ -48,12 +48,12 @@ struct child_sync { } \ } while (0) -#define PARENT_SKIP_IF_UNSUPPORTED (x, sync) \ +#define PARENT_SKIP_IF_UNSUPPORTED (x, sync, msg) \ do { \ if ((x) == -1 && (errno == ENODEV || errno == EINVAL)) { \ (sync)->parent_gave_up = true; \ prod_child(sync); \ - SKIP_IF(1); \ + SKIP_IF_MSG(1, msg); \ } \ } while (0)
tools/testing/selftests/powerpc/ptrace/core-pkey.c +1 -1
@@ -266,7 +266,7 @@ static int parent(struct shared_info *info, pid_t pid)* to the child. */ ret =ptrace_read_regs (pid,NT_PPC_PKEY , regs,3 ); -PARENT_SKIP_IF_UNSUPPORTED (ret, &info->child_sync); +PARENT_SKIP_IF_UNSUPPORTED (ret, &info->child_sync,"PKEYs not supported" );PARENT_FAIL_IF (ret, &info->child_sync); info->amr = regs[0 ];
tools/testing/selftests/powerpc/ptrace/perf-hwbreak.c +1 -1
@@ -884,7 +884,7 @@ static int perf_hwbreak(void) {srand (time (NULL ) ); -SKIP_IF (!perf_breakpoint_supported ()); +SKIP_IF_MSG (!perf_breakpoint_supported (),"Perf breakpoints not supported" );return runtest (); }
tools/testing/selftests/powerpc/ptrace/ptrace-hwbreak.c +1 -1
@@ -603,7 +603,7 @@ static int ptrace_hwbreak(void)wait (NULL );get_dbginfo (child_pid, &dbginfo); -SKIP_IF (dbginfo.num_data_bps ==0 ); +SKIP_IF_MSG (dbginfo.num_data_bps ==0 ,"No data breakpoints present" ); dawr =dawr_present (&dbginfo);run_tests (child_pid, &dbginfo, dawr);
tools/testing/selftests/powerpc/ptrace/ptrace-perf-hwbreak.c +2 -2
@@ -641,10 +641,10 @@ static int ptrace_perf_hwbreak(void)wait (NULL );/* <-- child (SIGUSR1) */ get_dbginfo (child_pid, &dbginfo); -SKIP_IF (dbginfo.num_data_bps <=1 ); +SKIP_IF_MSG (dbginfo.num_data_bps <=1 ,"Not enough data watchpoints (need at least 2)" ); ret =perf_cpu_event_open (0 , (__u64 )perf_data1,sizeof (*perf_data1)); -SKIP_IF (ret <0 ); +SKIP_IF_MSG (ret <0 ,"perf_event_open syscall failed" );close (ret); ret =test (child_pid);
tools/testing/selftests/powerpc/ptrace/ptrace-pkey.c +1 -1
@@ -192,7 +192,7 @@ static int parent(struct shared_info *info, pid_t pid)* to the child. */ ret =ptrace_read_regs (pid,NT_PPC_PKEY , regs,3 ); -PARENT_SKIP_IF_UNSUPPORTED (ret, &info->child_sync); +PARENT_SKIP_IF_UNSUPPORTED (ret, &info->child_sync,"PKEYs not supported" );PARENT_FAIL_IF (ret, &info->child_sync); info->amr1 = info->amr2 = regs[0 ];
tools/testing/selftests/powerpc/ptrace/ptrace-tar.c +1 -1
@@ -79,7 +79,7 @@ int ptrace_tar(void)int ret, status;// TAR was added in v2.07 -SKIP_IF (!have_hwcap2 (PPC_FEATURE2_ARCH_2_07 )); +SKIP_IF_MSG (!have_hwcap2 (PPC_FEATURE2_ARCH_2_07 ),"TAR requires ISA 2.07 compatible hardware" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *3 ,0777 |IPC_CREAT ); pid =fork ();
tools/testing/selftests/powerpc/ptrace/ptrace-tm-gpr.c +2 -2
@@ -112,8 +112,8 @@ int ptrace_tm_gpr(void)pid_t pid;int ret, status; -SKIP_IF (!have_htm ()); -SKIP_IF (htm_is_synthetic ()); +SKIP_IF_MSG (!have_htm (),"Don't have transactional memory" ); +SKIP_IF_MSG (htm_is_synthetic (),"Transactional memory is synthetic" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *2 ,0777 |IPC_CREAT ); pid =fork ();if (pid <0 ) {
tools/testing/selftests/powerpc/ptrace/ptrace-tm-spd-gpr.c +2 -2
@@ -118,8 +118,8 @@ int ptrace_tm_spd_gpr(void)pid_t pid;int ret, status; -SKIP_IF (!have_htm ()); -SKIP_IF (htm_is_synthetic ()); +SKIP_IF_MSG (!have_htm (),"Don't have transactional memory" ); +SKIP_IF_MSG (htm_is_synthetic (),"Transactional memory is synthetic" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *3 ,0777 |IPC_CREAT ); pid =fork ();if (pid <0 ) {
tools/testing/selftests/powerpc/ptrace/ptrace-tm-spd-tar.c +2 -2
@@ -128,8 +128,8 @@ int ptrace_tm_spd_tar(void)pid_t pid;int ret, status; -SKIP_IF (!have_htm ()); -SKIP_IF (htm_is_synthetic ()); +SKIP_IF_MSG (!have_htm (),"Don't have transactional memory" ); +SKIP_IF_MSG (htm_is_synthetic (),"Transactional memory is synthetic" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *3 ,0777 |IPC_CREAT ); pid =fork ();if (pid ==0 )
tools/testing/selftests/powerpc/ptrace/ptrace-tm-spd-vsx.c +2 -2
@@ -128,8 +128,8 @@ int ptrace_tm_spd_vsx(void)pid_t pid;int ret, status, i; -SKIP_IF (!have_htm ()); -SKIP_IF (htm_is_synthetic ()); +SKIP_IF_MSG (!have_htm (),"Don't have transactional memory" ); +SKIP_IF_MSG (htm_is_synthetic (),"Transactional memory is synthetic" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *3 ,0777 |IPC_CREAT );for (i =0 ; i <128 ; i++) {
tools/testing/selftests/powerpc/ptrace/ptrace-tm-spr.c +2 -2
@@ -113,8 +113,8 @@ int ptrace_tm_spr(void)pid_t pid;int ret, status; -SKIP_IF (!have_htm ()); -SKIP_IF (htm_is_synthetic ()); +SKIP_IF_MSG (!have_htm (),"Don't have transactional memory" ); +SKIP_IF_MSG (htm_is_synthetic (),"Transactional memory is synthetic" ); shm_id =shmget (IPC_PRIVATE ,sizeof (struct shared ),0777 |IPC_CREAT ); shm_id1 =shmget (IPC_PRIVATE ,sizeof (int ),0777 |IPC_CREAT ); pid =fork ();
tools/testing/selftests/powerpc/ptrace/ptrace-tm-tar.c +2 -2
@@ -116,8 +116,8 @@ int ptrace_tm_tar(void)pid_t pid;int ret, status; -SKIP_IF (!have_htm ()); -SKIP_IF (htm_is_synthetic ()); +SKIP_IF_MSG (!have_htm (),"Don't have transactional memory" ); +SKIP_IF_MSG (htm_is_synthetic (),"Transactional memory is synthetic" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *2 ,0777 |IPC_CREAT ); pid =fork ();if (pid ==0 )
tools/testing/selftests/powerpc/ptrace/ptrace-tm-vsx.c +2 -2
@@ -112,8 +112,8 @@ int ptrace_tm_vsx(void)pid_t pid;int ret, status, i; -SKIP_IF (!have_htm ()); -SKIP_IF (htm_is_synthetic ()); +SKIP_IF_MSG (!have_htm (),"Don't have transactional memory" ); +SKIP_IF_MSG (htm_is_synthetic (),"Transactional memory is synthetic" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *2 ,0777 |IPC_CREAT );for (i =0 ; i <128 ; i++) {
tools/testing/selftests/powerpc/ptrace/ptrace-vsx.c +1 -1
@@ -61,7 +61,7 @@ int ptrace_vsx(void)pid_t pid;int ret, status, i; -SKIP_IF (!have_hwcap (PPC_FEATURE_HAS_VSX )); +SKIP_IF_MSG (!have_hwcap (PPC_FEATURE_HAS_VSX ),"Don't have VSX" ); shm_id =shmget (IPC_PRIVATE ,sizeof (int ) *2 ,0777 |IPC_CREAT );
Signature · 1 line
2.41.0
PATCH 3/4 selftests/powerpc/ptrace: Fix typo in pid_max search error
In reply to Benjamin Graypid_max_addr() searches for the 'pid_max' symbol in /proc/kallsyms, and prints an error if it cannot find it. The error message has a typo, calling it pix_max.
--- tools/testing/selftests/powerpc/ptrace/ptrace-perf-hwbreak.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/powerpc/ptrace/ptrace-perf-hwbreak.c +1 -1
@@ -46,7 +46,7 @@ static unsigned long pid_max_addr(void)return strtoul (addr, &c,16 ); }fclose (fp); -printf ("Could not find pix_max. Exiting..\n" ); +printf ("Could not find pid_max. Exiting..\n" );exit (EXIT_FAILURE );return -1 ; }
Signature · 1 line
2.41.0
Michael Ellerman
AppliedRe: (subset) [PATCH v1 0/4] Improve ptrace selftest usability
In reply to Benjamin GrayQuoted 6 lines · Benjamin Gray
While trying to test changes to the breakpoint implementation in the kernel, I tried to run the ptrace tests and met many unexplained skips and failures.
This series addresses the pain points of trying to run these tests and learn what they are doing.
Quoted 7 lines
Benjamin Gray (4):
Documentation/powerpc: Fix ptrace request names selftests/powerpc/ptrace: Explain why tests are skipped selftests/powerpc/ptrace: Fix typo inpid_maxsearch error selftests/powerpc/ptrace: Declare test temporary variables as volatile[...]
Applied to powerpc/next.
[2/4] selftests/powerpc/ptrace: Explain why tests are skipped
https://git.kernel.org/powerpc/c/68877ff20a7f4f773069784cfe4f6fe9c7b9a841 [3/4] selftests/powerpc/ptrace: Fix typo in pid_max search error https://git.kernel.org/powerpc/c/fc6732a8556c1099b89f4777a96bd6a1ae5a4378 [4/4] selftests/powerpc/ptrace: Declare test temporary variables as volatile https://git.kernel.org/powerpc/c/c3062ede9927053754ba27b280afe00b9b31b37a
cheers
Re: (subset) [PATCH v1 0/4] Improve ptrace selftest usability
In reply to Benjamin GrayQuoted 6 lines · Benjamin Gray
While trying to test changes to the breakpoint implementation in the kernel, I tried to run the ptrace tests and met many unexplained skips and failures.
This series addresses the pain points of trying to run these tests and learn what they are doing.
Quoted 7 lines
Benjamin Gray (4):
Documentation/powerpc: Fix ptrace request names selftests/powerpc/ptrace: Explain why tests are skipped selftests/powerpc/ptrace: Fix typo inpid_maxsearch error selftests/powerpc/ptrace: Declare test temporary variables as volatile[...]
Patch 1 applied to powerpc/next.
[1/4] Documentation/powerpc: Fix ptrace request names
https://git.kernel.org/powerpc/c/0e216fa576e1cdea3913fe82b7283fdfb58c5c07
cheers