selftests: vm: add process_mrelease tests
Subsystems (2)
PATCH v2 1/1 selftests: vm: add process_mrelease tests
Introduce process_mrelease syscall sanity tests which include tests which expect to fail:
- process_mrelease with invalid pidfd and flags inputs
- process_mrelease on a live process with no pending signals and valid process_mrelease usage which is expected to succeed. Because process_mrelease has to be used against a process with a pending SIGKILL, it's possible that the process exits before process_mrelease gets called. In such cases we retry the test with a victim that allocates twice more memory up to 1GB. This would require the victim process to spend more time during exit and process_mrelease has a better chance of catching the process before it exits and succeeding.
On success the test reports the amount of memory the child had to allocate for reaping to succeed. Sample output:
Success reaping a child with 1MB of memory allocations
On failure the test reports the failure. Sample outputs:
All process_mrelease attempts failed!
process_mrelease: Invalid argument
--- tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 1 + tools/testing/selftests/vm/mrelease_test.c | 214 +++++++++++++++++++++ tools/testing/selftests/vm/run_vmtests.sh | 16 ++ 4 files changed, 232 insertions(+) create mode 100644 tools/testing/selftests/vm/mrelease_test.c
tools/testing/selftests/vm/.gitignore +1 -0
@@ -10,6 +10,7 @@ map_populate thuge-gen compaction_test mlock2-tests +mrelease_test mremap_dontunmap mremap_test on-fault-limit
tools/testing/selftests/vm/Makefile +1 -0
@@ -43,6 +43,7 @@ TEST_GEN_FILES += map_populate TEST_GEN_FILES += memfd_secret TEST_GEN_FILES += mlock-random-test TEST_GEN_FILES += mlock2-tests +TEST_GEN_FILES += mrelease_test TEST_GEN_FILES += mremap_dontunmap TEST_GEN_FILES += mremap_test TEST_GEN_FILES += on-fault-limit
tools/testing/selftests/vm/mrelease_test.c new file +214 -0
214 lines not on this page · raw message
tools/testing/selftests/vm/run_vmtests.sh +16 -0
@@ -287,6 +287,22 @@ else echo "[PASS]" fi +echo "---------------------" +echo "running mrelease_test" +echo "---------------------" +./mrelease_test +ret_val=$? + +if [ $ret_val -eq 0 ]; then + echo "[PASS]" +elif [ $ret_val -eq $ksft_skip ]; then + echo "[SKIP]" + exitcode=$ksft_skip +else + echo "[FAIL]" + exitcode=1 +fi + echo "-------------------" echo "running mremap_test" echo "-------------------"
Signature · 1 line
2.36.0.550.gb090851708-goog
Andrew Morton
maintainerQuoted 37 lines · On Mon, 16 May 2022 00:55:38 -0700 Suren Baghdasaryan <[email protected]> wrote:
Introduce
process_mreleasesyscall sanity tests which include tests which expect to fail:
-process_mreleasewith invalid pidfd and flags inputs
-process_mreleaseon a live process with no pending signals and validprocess_mreleaseusage which is expected to succeed. Becauseprocess_mreleasehas to be used against a process with a pending SIGKILL, it's possible that the process exits beforeprocess_mreleasegets called. In such cases we retry the test with a victim that allocates twice more memory up to 1GB. This would require the victim process to spend more time during exit andprocess_mreleasehas a better chance of catching the process before it exits and succeeding.On success the test reports the amount of memory the child had to allocate for reaping to succeed. Sample output:
Success reaping a child with 1MB of memory allocationsOn failure the test reports the failure. Sample outputs:
Allprocess_mreleaseattempts failed!
process_mrelease: Invalid argument...
tools/testing/selftests/vm/run_vmtests.sh +9
-0@@ -287,6 +287,22 @@ else echo "[PASS]" fi +echo "---------------------" +echo "running mrelease_test" +echo "---------------------" +./mrelease_test +ret_val=$? + +if [ $ret_val -eq 0 ]; then + echo "[PASS]" +elif [ $ret_val -eq $ksft_skip ]; then
tools/testing/selftests/vm/run_vmtests.sh else · line 308
Quoted 10 lines
tools/testing/selftests/vm/run_vmtests.sh +7
-0+ echo "[SKIP]" + exitcode=$ksft_skip +else + echo "[FAIL]" + exitcode=1 +fi + echo "-------------------" echo "running mremap_test" echo "-------------------"
Can you please redo this againstgit://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm branch mm-stable or mm-unstable. Or against linux-next?
This script now has a helper function run_test which I think can be used here.
Quoted 55 lines · Andrew Morton
Quoted 37 lines · On Mon, 16 May 2022 00:55:38 -0700 Suren Baghdasaryan <[email protected]> wrote:
Introduce
process_mreleasesyscall sanity tests which include tests which expect to fail:
-process_mreleasewith invalid pidfd and flags inputs
-process_mreleaseon a live process with no pending signals and validprocess_mreleaseusage which is expected to succeed. Becauseprocess_mreleasehas to be used against a process with a pending SIGKILL, it's possible that the process exits beforeprocess_mreleasegets called. In such cases we retry the test with a victim that allocates twice more memory up to 1GB. This would require the victim process to spend more time during exit andprocess_mreleasehas a better chance of catching the process before it exits and succeeding.On success the test reports the amount of memory the child had to allocate for reaping to succeed. Sample output:
Success reaping a child with 1MB of memory allocationsOn failure the test reports the failure. Sample outputs:
Allprocess_mreleaseattempts failed!
process_mrelease: Invalid argument...
tools/testing/selftests/vm/run_vmtests.sh +9
-0@@ -287,6 +287,22 @@ else echo "[PASS]" fi +echo "---------------------" +echo "running mrelease_test" +echo "---------------------" +./mrelease_test +ret_val=$? + +if [ $ret_val -eq 0 ]; then + echo "[PASS]" +elif [ $ret_val -eq $ksft_skip ]; thentools/testing/selftests/vm/run_vmtests.sh else · line 308
Quoted 10 lines
tools/testing/selftests/vm/run_vmtests.sh +7
-0+ echo "[SKIP]" + exitcode=$ksft_skip +else + echo "[FAIL]" + exitcode=1 +fi + echo "-------------------" echo "running mremap_test" echo "-------------------"Can you please redo this against
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mmbranch mm-stable or mm-unstable. Or against linux-next?
Quoted 2 lines
This script now has a helper function
run_testwhich I think can be used here.
Ah, thanks for the note! I'll give it a couple days for more reviews and the will post it over mmotm.
Shuah Khan
maintainerQuoted 12 lines · Suren Baghdasaryan
Introduce
process_mreleasesyscall sanity tests which include tests which expect to fail:
-process_mreleasewith invalid pidfd and flags inputs
-process_mreleaseon a live process with no pending signals and validprocess_mreleaseusage which is expected to succeed. Becauseprocess_mreleasehas to be used against a process with a pending SIGKILL, it's possible that the process exits beforeprocess_mreleasegets called. In such cases we retry the test with a victim that allocates twice more memory up to 1GB. This would require the victim process to spend more time during exit andprocess_mreleasehas a better chance of catching the process before it exits and succeeding.
Quoted 8 lines
On success the test reports the amount of memory the child had to allocate for reaping to succeed. Sample output:
Success reaping a child with 1MB of memory allocationsOn failure the test reports the failure. Sample outputs:
Allprocess_mreleaseattempts failed!
process_mrelease: Invalid argument
Nit: Please format this better - include actual example output from the command and how to run the test examples.
Quoted 65 lines
Signed-off-by Suren Baghdasaryan <[email protected]>--- tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 1 + tools/testing/selftests/vm/mrelease_test.c | 214 +++++++++++++++++++++ tools/testing/selftests/vm/run_vmtests.sh | 16 ++ 4 files changed, 232 insertions(+)create mode 100644
tools/testing/selftests/vm/mrelease_test.ctools/testing/selftests/vm/.gitignore +1
-0@@ -10,6 +10,7 @@ map_populate thuge-gen compaction_test mlock2-tests +mrelease_test mremap_dontunmap mremap_test on-fault-limittools/testing/selftests/vm/Makefile +1
-0@@ -43,6 +43,7 @@ TEST_GEN_FILES += map_populate TEST_GEN_FILES += memfd_secret TEST_GEN_FILES += mlock-random-test TEST_GEN_FILES += mlock2-tests +TEST_GEN_FILES += mrelease_test TEST_GEN_FILES += mremap_dontunmap TEST_GEN_FILES += mremap_test TEST_GEN_FILES += on-fault-limittools/testing/selftests/vm/mrelease_test.c new file +26
-0@@ -0,0 +1,214 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Copyright 2022 Google LLC +*/ +#define _GNU_SOURCE +#include <errno.h> +#include <stdbool.h> +#include <stdio.h> +#include <stdlib.h> +#include <sys/wait.h> +#include <unistd.h> + +#include "util.h" + +#include "../kselftest.h" + +#if defined(__NR_pidfd_open) && defined(__NR_process_mrelease) + +static inline int pidfd_open (pid_t pid,unsigned int flags) +{ +return syscall (__NR_pidfd_open, pid, flags); +} + +static inline int process_mrelease (int pidfd,unsigned int flags) +{ +return syscall (__NR_process_mrelease, pidfd, flags);
tools/testing/selftests/vm/mrelease_test.c line 36
Quoted 10 lines
tools/testing/selftests/vm/mrelease_test.c new file +10
-0+} + +static void write_fault_pages (char *addr,unsigned long nr_pages) +{ +unsigned long i; + +for (i =0 ; i < nr_pages; i++) + *((unsigned long *)(addr + (i *PAGE_SIZE ))) = i; +} +
Okay these above 3 routines are called once. I am not seeing any point in making these separate routines. I made the same comment on v1.
Quoted 88 lines
tools/testing/selftests/vm/mrelease_test.c new file +88
-0+static int alloc_noexit (unsigned long nr_pages,int pipefd) +{ +int timeout =10 ;/* 10sec timeout to get killed */ +int ppid =getppid (); +void *buf; + + buf =mmap (NULL , nr_pages *PAGE_SIZE ,PROT_READ |PROT_WRITE , +MAP_PRIVATE |MAP_ANON ,0 ,0 ); +if (buf ==MAP_FAILED ) { +perror ("mmap failed, halting the test" ); +return KSFT_FAIL ; + } + +write_fault_pages ((char *)buf, nr_pages); + +/* Signal the parent that the child is ready */ +if (write (pipefd,"" ,1 ) <0 ) { +perror ("write" ); +return KSFT_FAIL ; + } + +/* Wait to be killed (when reparenting happens) */ +while (getppid () == ppid && timeout >0 ) { +sleep (1 ); + timeout--; + } + +munmap (buf, nr_pages *PAGE_SIZE ); + +return (timeout >0 ) ?KSFT_PASS :KSFT_FAIL ; +} + +/* The process_mrelease calls in this test are expected to fail */ +void run_negative_tests (int pidfd) +{ +/* Test invalid flags. Expect to fail with EINVAL error code. */ +if (!process_mrelease (pidfd, (unsigned int )-1 ) || errno !=EINVAL ) { +perror ("process_mrelease with wrong flags" ); +exit (KSFT_FAIL ); + } +/* +* Test reaping while process is alive with no pending SIGKILL. +* Expect to fail with EINVAL error code. +*/ +if (!process_mrelease (pidfd,0 ) || errno !=EINVAL ) { +perror ("process_mrelease on a live process" ); +exit (KSFT_FAIL ); + } +} + +#define MB (x) (x << 20) +#define MAX_SIZE_MB 1024 + +int main (void ) +{ +int pipefd[2 ], pidfd; +bool success, retry; +size_t size; +pid_t pid; +char byte; +int res; + +/* Test a wrong pidfd */ +if (!process_mrelease (-1 ,0 ) || errno !=EBADF ) { +perror ("process_mrelease with wrong pidfd" ); +exit (KSFT_FAIL ); + } + +/* Start the test with 1MB child memory allocation */ + size =1 ; +retry: +/* +* Pipe for the child to signal when it's done allocating +* memory +*/ +if (pipe (pipefd)) { +perror ("pipe" ); +exit (KSFT_FAIL ); + } + pid =fork (); +if (pid <0 ) { +perror ("fork" ); +close (pipefd[0 ]); +close (pipefd[1 ]); +exit (KSFT_FAIL ); + } + +if (pid ==0 ) {
tools/testing/selftests/vm/mrelease_test.c line 134
Quoted 10 lines
tools/testing/selftests/vm/mrelease_test.c new file +10
-0+/* +* Child main routine: +* Allocate and fault-in memory and wait to be killed +*/ +close (pipefd[0 ]); + res =alloc_noexit (MB (size) /PAGE_SIZE , pipefd[1 ]); +close (pipefd[1 ]); +exit (res); + } +
Now the above code can be a separate function which will make it readable.
Quoted 69 lines
tools/testing/selftests/vm/mrelease_test.c new file +69
-0+/* +* Parent main routine: +* Wait for the child to finish allocations, then kill and reap +*/ +close (pipefd[1 ]); +/* Block until the child is ready */ + res =read (pipefd[0 ], &byte,1 ); +close (pipefd[0 ]); +if (res <0 ) { +perror ("read" ); +if (!kill (pid,SIGKILL )) +waitpid (pid,NULL ,0 ); +exit (KSFT_FAIL ); + } + + pidfd =pidfd_open (pid,0 ); +if (pidfd <0 ) { +perror ("pidfd_open" ); +if (!kill (pid,SIGKILL )) +waitpid (pid,NULL ,0 ); +exit (KSFT_FAIL ); + } + +/* Run negative tests which require a live child */ +run_negative_tests (pidfd); + +if (kill (pid,SIGKILL )) { +perror ("kill" ); +exit (KSFT_FAIL ); + } + + success = (process_mrelease (pidfd,0 ) ==0 ); +if (!success) { +/* +* If we failed to reap because the child exited too soon, +* before we could call process_mrelease. Double child's memory +* which causes it to spend more time on cleanup and increases +* our chances of reaping its memory before it exits. +* Retry until we succeed or reach MAX_SIZE_MB. +*/ +if (errno ==ESRCH ) { + retry = (size <=MAX_SIZE_MB ); + }else { +perror ("process_mrelease" ); +waitpid (pid,NULL ,0 ); +exit (KSFT_FAIL ); + } + } + +/* Cleanup to prevent zombies */ +if (waitpid (pid,NULL ,0 ) <0 ) { +perror ("waitpid" ); +exit (KSFT_FAIL ); + } +close (pidfd); + +if (!success) { +if (retry) { + size *=2 ; + goto retry; + } +printf ("All process_mrelease attempts failed!\n" ); +exit (KSFT_FAIL ); + } + +printf ("Success reaping a child with %zuMB of memory allocations\n" , + size); +return KSFT_PASS ; +}
tools/testing/selftests/vm/mrelease_test.c line 213
Quoted 10 lines
tools/testing/selftests/vm/mrelease_test.c new file +10
-0+ +#else /* defined(__NR_pidfd_open) && defined(__NR_process_mrelease) */ + +int main (int argc,char *argv[]) +{ +printf ("skip: skipping process_mrelease test " \ +"(missing __NR_pidfd_open and/or __NR_process_mrelease)\n" ); +return KSFT_SKIP ; +} +
Why do you need these ifdefs - syscall will return ENOSYS and you can key off that. Please take a look at other usages of syscall in the repo.
Quoted 19 lines
tools/testing/selftests/vm/mrelease_test.c new file +1
-0+#endif /* defined(__NR_pidfd_open) && defined(__NR_process_mrelease) */ tools/testing/selftests/vm/run_vmtests.sh +9
-0@@ -287,6 +287,22 @@ else echo "[PASS]" fi +echo "---------------------" +echo "running mrelease_test" +echo "---------------------" +./mrelease_test +ret_val=$? + +if [ $ret_val -eq 0 ]; then + echo "[PASS]" +elif [ $ret_val -eq $ksft_skip ]; then
tools/testing/selftests/vm/run_vmtests.sh else · line 308
Quoted 10 lines
tools/testing/selftests/vm/run_vmtests.sh +7
-0+ echo "[SKIP]" + exitcode=$ksft_skip +else + echo "[FAIL]" + exitcode=1 +fi + echo "-------------------" echo "running mremap_test" echo "-------------------"
thanks,
-- Shuah
Quoted 18 lines · Shuah Khan
Quoted 16 lines · Suren Baghdasaryan
Introduce
process_mreleasesyscall sanity tests which include tests which expect to fail:
-process_mreleasewith invalid pidfd and flags inputs
-process_mreleaseon a live process with no pending signals and validprocess_mreleaseusage which is expected to succeed. Becauseprocess_mreleasehas to be used against a process with a pending SIGKILL, it's possible that the process exits beforeprocess_mreleasegets called. In such cases we retry the test with a victim that allocates twice more memory up to 1GB. This would require the victim process to spend more time during exit andprocess_mreleasehas a better chance of catching the process before it exits and succeeding.On success the test reports the amount of memory the child had to allocate for reaping to succeed. Sample output:
Success reaping a child with 1MB of memory allocations
Quoted 7 lines
Quoted 4 lines
On failure the test reports the failure. Sample outputs:
Allprocess_mreleaseattempts failed!
process_mrelease: Invalid argumentNit: Please format this better - include actual example output from the command and how to run the test examples.
Hmm... Those are the actual outputs from the command and it does not take any input arguments. Do you mean smth like this:
$ mrelease_test Success reaping a child with 1MB of memory allocations
$ mrelease_test All process_mrelease attempts failed!
$ mrelease_test process_mrelease: Invalid argument
?
Quoted 71 lines
Quoted 71 lines
Signed-off-by Suren Baghdasaryan <[email protected]>--- tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 1 + tools/testing/selftests/vm/mrelease_test.c | 214 +++++++++++++++++++++ tools/testing/selftests/vm/run_vmtests.sh | 16 ++ 4 files changed, 232 insertions(+)create mode 100644
tools/testing/selftests/vm/mrelease_test.ctools/testing/selftests/vm/.gitignore +1
-0@@ -10,6 +10,7 @@ map_populate thuge-gen compaction_test mlock2-tests +mrelease_test mremap_dontunmap mremap_test on-fault-limittools/testing/selftests/vm/Makefile +1
-0@@ -43,6 +43,7 @@ TEST_GEN_FILES += map_populate TEST_GEN_FILES += memfd_secret TEST_GEN_FILES += mlock-random-test TEST_GEN_FILES += mlock2-tests +TEST_GEN_FILES += mrelease_test TEST_GEN_FILES += mremap_dontunmap TEST_GEN_FILES += mremap_test TEST_GEN_FILES += on-fault-limittools/testing/selftests/vm/mrelease_test.c new file +31
-0@@ -0,0 +1,214 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Copyright 2022 Google LLC +*/ +#define _GNU_SOURCE +#include <errno.h> +#include <stdbool.h> +#include <stdio.h> +#include <stdlib.h> +#include <sys/wait.h> +#include <unistd.h> + +#include "util.h" + +#include "../kselftest.h" + +#if defined(__NR_pidfd_open) && defined(__NR_process_mrelease) + +static inline int pidfd_open (pid_t pid,unsigned int flags) +{ +return syscall (__NR_pidfd_open, pid, flags); +} + +static inline int process_mrelease (int pidfd,unsigned int flags) +{ +return syscall (__NR_process_mrelease, pidfd, flags); +} + +static void write_fault_pages (char *addr,unsigned long nr_pages) +{ +unsigned long i;
Quoted 8 lines
tools/testing/selftests/vm/mrelease_test.c line 36
Quoted 5 lines
tools/testing/selftests/vm/mrelease_test.c new file +5
-0+ +for (i =0 ; i < nr_pages; i++) + *((unsigned long *)(addr + (i *PAGE_SIZE ))) = i; +} +Okay these above 3 routines are called once. I am not seeing any point in making these separate routines. I made the same comment on v1.
I must have misunderstood your previous comment. Will change.
Quoted 93 lines
Quoted 93 lines
tools/testing/selftests/vm/mrelease_test.c new file +92
-0+static int alloc_noexit (unsigned long nr_pages,int pipefd) +{ +int timeout =10 ;/* 10sec timeout to get killed */ +int ppid =getppid (); +void *buf; + + buf =mmap (NULL , nr_pages *PAGE_SIZE ,PROT_READ |PROT_WRITE , +MAP_PRIVATE |MAP_ANON ,0 ,0 ); +if (buf ==MAP_FAILED ) { +perror ("mmap failed, halting the test" ); +return KSFT_FAIL ; + } + +write_fault_pages ((char *)buf, nr_pages); + +/* Signal the parent that the child is ready */ +if (write (pipefd,"" ,1 ) <0 ) { +perror ("write" ); +return KSFT_FAIL ; + } + +/* Wait to be killed (when reparenting happens) */ +while (getppid () == ppid && timeout >0 ) { +sleep (1 ); + timeout--; + } + +munmap (buf, nr_pages *PAGE_SIZE ); + +return (timeout >0 ) ?KSFT_PASS :KSFT_FAIL ; +} + +/* The process_mrelease calls in this test are expected to fail */ +void run_negative_tests (int pidfd) +{ +/* Test invalid flags. Expect to fail with EINVAL error code. */ +if (!process_mrelease (pidfd, (unsigned int )-1 ) || errno !=EINVAL ) { +perror ("process_mrelease with wrong flags" ); +exit (KSFT_FAIL ); + } +/* +* Test reaping while process is alive with no pending SIGKILL. +* Expect to fail with EINVAL error code. +*/ +if (!process_mrelease (pidfd,0 ) || errno !=EINVAL ) { +perror ("process_mrelease on a live process" ); +exit (KSFT_FAIL ); + } +} + +#define MB (x) (x << 20) +#define MAX_SIZE_MB 1024 + +int main (void ) +{ +int pipefd[2 ], pidfd; +bool success, retry; +size_t size; +pid_t pid; +char byte; +int res; + +/* Test a wrong pidfd */ +if (!process_mrelease (-1 ,0 ) || errno !=EBADF ) { +perror ("process_mrelease with wrong pidfd" ); +exit (KSFT_FAIL ); + } + +/* Start the test with 1MB child memory allocation */ + size =1 ; +retry: +/* +* Pipe for the child to signal when it's done allocating +* memory +*/ +if (pipe (pipefd)) { +perror ("pipe" ); +exit (KSFT_FAIL ); + } + pid =fork (); +if (pid <0 ) { +perror ("fork" ); +close (pipefd[0 ]); +close (pipefd[1 ]); +exit (KSFT_FAIL ); + } + +if (pid ==0 ) { +/* +* Child main routine: +* Allocate and fault-in memory and wait to be killed +*/
Quoted 8 lines
tools/testing/selftests/vm/mrelease_test.c line 134
Quoted 6 lines
tools/testing/selftests/vm/mrelease_test.c new file +6
-0+close (pipefd[0 ]); + res =alloc_noexit (MB (size) /PAGE_SIZE , pipefd[1 ]); +close (pipefd[1 ]); +exit (res); + } +Now the above code can be a separate function which will make it readable.
Ack.
Quoted 76 lines
Quoted 76 lines
tools/testing/selftests/vm/mrelease_test.c new file +75
-0+/* +* Parent main routine: +* Wait for the child to finish allocations, then kill and reap +*/ +close (pipefd[1 ]); +/* Block until the child is ready */ + res =read (pipefd[0 ], &byte,1 ); +close (pipefd[0 ]); +if (res <0 ) { +perror ("read" ); +if (!kill (pid,SIGKILL )) +waitpid (pid,NULL ,0 ); +exit (KSFT_FAIL ); + } + + pidfd =pidfd_open (pid,0 ); +if (pidfd <0 ) { +perror ("pidfd_open" ); +if (!kill (pid,SIGKILL )) +waitpid (pid,NULL ,0 ); +exit (KSFT_FAIL ); + } + +/* Run negative tests which require a live child */ +run_negative_tests (pidfd); + +if (kill (pid,SIGKILL )) { +perror ("kill" ); +exit (KSFT_FAIL ); + } + + success = (process_mrelease (pidfd,0 ) ==0 ); +if (!success) { +/* +* If we failed to reap because the child exited too soon, +* before we could call process_mrelease. Double child's memory +* which causes it to spend more time on cleanup and increases +* our chances of reaping its memory before it exits. +* Retry until we succeed or reach MAX_SIZE_MB. +*/ +if (errno ==ESRCH ) { + retry = (size <=MAX_SIZE_MB ); + }else { +perror ("process_mrelease" ); +waitpid (pid,NULL ,0 ); +exit (KSFT_FAIL ); + } + } + +/* Cleanup to prevent zombies */ +if (waitpid (pid,NULL ,0 ) <0 ) { +perror ("waitpid" ); +exit (KSFT_FAIL ); + } +close (pidfd); + +if (!success) { +if (retry) { + size *=2 ; + goto retry; + } +printf ("All process_mrelease attempts failed!\n" ); +exit (KSFT_FAIL ); + } + +printf ("Success reaping a child with %zuMB of memory allocations\n" , + size); +return KSFT_PASS ; +} + +#else /* defined(__NR_pidfd_open) && defined(__NR_process_mrelease) */ + +int main (int argc,char *argv[]) +{ + printf("skip: skipping process_mrelease test " \
Quoted 8 lines
tools/testing/selftests/vm/mrelease_test.c line 213
Quoted 4 lines
tools/testing/selftests/vm/mrelease_test.c new file +4
-0+ "(missing __NR_pidfd_open and/or __NR_process_mrelease)\n"); +return KSFT_SKIP ; +} +Why do you need these ifdefs - syscall will return ENOSYS and you can key off that. Please take a look at other usages of syscall in the repo.
The issue is that I need to provide the syscall number when callingsyscall() (in my case __NR_pidfd_open and __NR_process_mrelease) and
if that number is not defined in the userspace headers on a given
system then what should I pass instead?
When implementing this I followed the examples of
https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/memfd_secret.c#L30
and https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/userfaultfd.c#L65.
My original implementation was modeled after this approach:
https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/mlock2.h#L15.
If none of these are correct, could you please point me to the example
you want me to follow?
Quoted 33 lines
Quoted 20 lines
tools/testing/selftests/vm/mrelease_test.c new file +1
-0+#endif /* defined(__NR_pidfd_open) && defined(__NR_process_mrelease) */ tools/testing/selftests/vm/run_vmtests.sh +9
-0@@ -287,6 +287,22 @@ else echo "[PASS]" fi +echo "---------------------" +echo "running mrelease_test" +echo "---------------------" +./mrelease_test +ret_val=$? + +if [ $ret_val -eq 0 ]; then + echo "[PASS]" +elif [ $ret_val -eq $ksft_skip ]; thentools/testing/selftests/vm/run_vmtests.sh else · line 308
Quoted 10 lines
tools/testing/selftests/vm/run_vmtests.sh +7
-0+ echo "[SKIP]" + exitcode=$ksft_skip +else + echo "[FAIL]" + exitcode=1 +fi + echo "-------------------" echo "running mremap_test" echo "-------------------"thanks,
-- Shuah
Shuah Khan
maintainerQuoted 33 lines · Suren Baghdasaryan
Quoted 18 lines · Shuah Khan
Quoted 16 lines · Suren Baghdasaryan
Introduce
process_mreleasesyscall sanity tests which include tests which expect to fail:
-process_mreleasewith invalid pidfd and flags inputs
-process_mreleaseon a live process with no pending signals and validprocess_mreleaseusage which is expected to succeed. Becauseprocess_mreleasehas to be used against a process with a pending SIGKILL, it's possible that the process exits beforeprocess_mreleasegets called. In such cases we retry the test with a victim that allocates twice more memory up to 1GB. This would require the victim process to spend more time during exit andprocess_mreleasehas a better chance of catching the process before it exits and succeeding.On success the test reports the amount of memory the child had to allocate for reaping to succeed. Sample output:
Success reaping a child with 1MB of memory allocationsQuoted 7 lines
Quoted 4 lines
On failure the test reports the failure. Sample outputs:
Allprocess_mreleaseattempts failed!
process_mrelease: Invalid argumentNit: Please format this better - include actual example output from the command and how to run the test examples.
Hmm... Those are the actual outputs from the command and it does not take any input arguments. Do you mean smth like this:
$ mrelease_test Success reaping a child with 1MB of memory allocations
Quoted 7 lines
$ mrelease_test All process_mrelease attempts failed!$ mrelease_test process_mrelease: Invalid argument?
This looks good.
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Signed-off-by Suren Baghdasaryan <[email protected]>--- tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 1 + tools/testing/selftests/vm/mrelease_test.c | 214 +++++++++++++++++++++ tools/testing/selftests/vm/run_vmtests.sh | 16 ++ 4 files changed, 232 insertions(+)create mode 100644
tools/testing/selftests/vm/mrelease_test.c
[snip]
Quoted 6 lines
Quoted 3 lines
Okay these above 3 routines are called once. I am not seeing any point in making these separate routines. I made the same comment on v1.
I must have misunderstood your previous comment. Will change.
Thank you.
Quoted 15 lines
Quoted 4 lines
Now the above code can be a separate function which will make it readable.
Ack.
Quoted 7 lines
Quoted 3 lines
+
Why do you need these ifdefs - syscall will return ENOSYS and you can key off that. Please take a look at other usages of syscall in the repo.
Quoted 12 lines
The issue is that I need to provide the syscall number when calling
syscall()(in my case__NR_pidfd_openand__NR_process_mrelease) and
if that number is not defined in the userspace headers on a given
system then what should I pass instead?
When implementing this I followed the examples of
https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/memfd_secret.c#L30
and https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/userfaultfd.c#L65.
My original implementation was modeled after this approach:
https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/mlock2.h#L15.
If none of these are correct, could you please point me to the example
you want me to follow?
kselftests include kernel headers. As long as these syscalls are defined in the kernel headers, the test will build.
Looks it is defined in include/uapi/asm-generic/unistd.h
You can assume it is defined and then if we find architectures that don't, you can follow what tools/testing/selftests/pidfd/pidfd.h does.
This way the test can simply call syscall and handle ENOSYS.
thanks,
-- Shuah
Quoted 99 lines · Shuah Khan
Quoted 33 lines · Suren Baghdasaryan
Quoted 18 lines · Shuah Khan
Quoted 16 lines · Suren Baghdasaryan
Introduce
process_mreleasesyscall sanity tests which include tests which expect to fail:
-process_mreleasewith invalid pidfd and flags inputs
-process_mreleaseon a live process with no pending signals and validprocess_mreleaseusage which is expected to succeed. Becauseprocess_mreleasehas to be used against a process with a pending SIGKILL, it's possible that the process exits beforeprocess_mreleasegets called. In such cases we retry the test with a victim that allocates twice more memory up to 1GB. This would require the victim process to spend more time during exit andprocess_mreleasehas a better chance of catching the process before it exits and succeeding.On success the test reports the amount of memory the child had to allocate for reaping to succeed. Sample output:
Success reaping a child with 1MB of memory allocationsQuoted 7 lines
Quoted 4 lines
On failure the test reports the failure. Sample outputs:
Allprocess_mreleaseattempts failed!
process_mrelease: Invalid argumentNit: Please format this better - include actual example output from the command and how to run the test examples.
Hmm... Those are the actual outputs from the command and it does not take any input arguments. Do you mean smth like this:
$ mrelease_test Success reaping a child with 1MB of memory allocationsQuoted 7 lines
$ mrelease_test All process_mrelease attempts failed!$ mrelease_test process_mrelease: Invalid argument?
This looks good.
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Signed-off-by Suren Baghdasaryan <[email protected]>--- tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 1 + tools/testing/selftests/vm/mrelease_test.c | 214 +++++++++++++++++++++ tools/testing/selftests/vm/run_vmtests.sh | 16 ++ 4 files changed, 232 insertions(+)create mode 100644
tools/testing/selftests/vm/mrelease_test.c[snip]
Quoted 6 lines
Quoted 3 lines
Okay these above 3 routines are called once. I am not seeing any point in making these separate routines. I made the same comment on v1.
I must have misunderstood your previous comment. Will change.
Thank you.
Quoted 15 lines
Quoted 4 lines
Now the above code can be a separate function which will make it readable.
Ack.
Quoted 7 lines
Quoted 3 lines
+
Why do you need these ifdefs - syscall will return ENOSYS and you can key off that. Please take a look at other usages of syscall in the repo.
Quoted 12 lines
The issue is that I need to provide the syscall number when calling
syscall()(in my case__NR_pidfd_openand__NR_process_mrelease) and
if that number is not defined in the userspace headers on a given
system then what should I pass instead?
When implementing this I followed the examples of
https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/memfd_secret.c#L30
and https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/userfaultfd.c#L65.
My original implementation was modeled after this approach:
https://elixir.bootlin.com/linux/latest/source/tools/testing/selftests/vm/mlock2.h#L15.
If none of these are correct, could you please point me to the example
you want me to follow?kselftests include kernel headers. As long as these syscalls are defined in the kernel headers, the test will build.
Quoted 7 lines
Looks it is defined in
include/uapi/asm-generic/unistd.hYou can assume it is defined and then if we find architectures that don't, you can follow what
tools/testing/selftests/pidfd/pidfd.hdoes.This way the test can simply call syscall and handle ENOSYS.
Thanks for the guidance! I'll try that approach.
Suren.
Quoted 10 lines
thanks,
-- Shuah--
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].