Message ID | 20200609095102.21153-1-liwang@redhat.com |
---|---|
State | Accepted |
Headers | show |
Series | [v4,1/4] lib: add new cgroup test API | expand |
----- Original Message ----- > Many of our LTP tests need Control Group in the configuration, > this patch makes cgroup unified mounting at setup phase to be > possible. The method is extracted from mem.h with the purpose > of extendible for further developing, and trying to compatible > the current two versions of cgroup, > > It's hard to make all APIs be strictly consistent because there > are many differences between v1 and v2. But it capsulate the detail > of cgroup mounting in high-level functions, which will be easier > to use cgroup without considering too much technical thing. > > Btw, test get passed on RHEL7(x86_64), RHEL8(ppc64le), Fedora32(x86_64). > No strong objections to v4, couple comments below (if you spin v5 because of other reviews). > +2.2.36 Using Control Group > +^^^^^^^^^^^^^^^^^^^^^^^^^^ Would be nice if there was short description of each function. > +static void tst_cgroup_set_path(const char *cgroup_dir) > +{ > + char cgroup_new_dir[PATH_MAX]; > + struct tst_cgroup_path *tst_cgroup_path, *a; > + > + if (!cgroup_dir) > + tst_brk(TBROK, "Invalid cgroup dir, plese check cgroup_dir"); > + > + sprintf(cgroup_new_dir, "%s/ltp_%d", cgroup_dir, rand()); > + > + /* To store cgroup path in the 'path' list */ > + tst_cgroup_path = SAFE_MMAP(NULL, (sizeof(struct tst_cgroup_path)), > + PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); This looks like it could use just SAFE_MALLOC/SAFE_FREE.
On Tue, Jun 9, 2020 at 6:42 PM Jan Stancek <jstancek@redhat.com> wrote: > > ----- Original Message ----- > > Many of our LTP tests need Control Group in the configuration, > > this patch makes cgroup unified mounting at setup phase to be > > possible. The method is extracted from mem.h with the purpose > > of extendible for further developing, and trying to compatible > > the current two versions of cgroup, > > > > It's hard to make all APIs be strictly consistent because there > > are many differences between v1 and v2. But it capsulate the detail > > of cgroup mounting in high-level functions, which will be easier > > to use cgroup without considering too much technical thing. > > > > Btw, test get passed on RHEL7(x86_64), RHEL8(ppc64le), Fedora32(x86_64). > > > > No strong objections to v4, couple comments below (if you spin v5 because > of other reviews). > Thanks for review! Ok, I agree to keep patchv4 posts for more days in case other people have comments. > > > +2.2.36 Using Control Group > > +^^^^^^^^^^^^^^^^^^^^^^^^^^ > > Would be nice if there was short description of each function. > I'd add some code comments in the tst_cgroup.h header file. Which something maybe like: /* To mount/umount specified cgroup controller on 'cgroup_dir' path */ void tst_cgroup_mount(enum tst_cgroup_ctrl ctrl, const char *cgroup_dir); void tst_cgroup_umount(const char *cgroup_dir); /* To move current process PID to the mounted cgroup tasks */ void tst_cgroup_move_current(const char *cgroup_dir); /* To set cgroup controller knob with new value */ void tst_cgroup_set_knob(const char *cgroup_dir, const char *knob, long value); /* Set of functions to set knobs under the memory controller */ void tst_cgroup_mem_set_maxbytes(const char *cgroup_dir, long memsz); 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); > > > +static void tst_cgroup_set_path(const char *cgroup_dir) > > +{ > > + char cgroup_new_dir[PATH_MAX]; > > + struct tst_cgroup_path *tst_cgroup_path, *a; > > + > > + if (!cgroup_dir) > > + tst_brk(TBROK, "Invalid cgroup dir, plese check > cgroup_dir"); > > + > > + sprintf(cgroup_new_dir, "%s/ltp_%d", cgroup_dir, rand()); > > + > > + /* To store cgroup path in the 'path' list */ > > + tst_cgroup_path = SAFE_MMAP(NULL, (sizeof(struct tst_cgroup_path)), > > + PROT_READ | PROT_WRITE, MAP_PRIVATE | > MAP_ANONYMOUS, -1, 0); > > This looks like it could use just SAFE_MALLOC/SAFE_FREE. > Agree. And btw, seems we have to set cgroup.clone_children as 1, otherwise, we can't write anything to the cpuset subsystem files. BROK: Failed to close FILE '/tmp/cgroup_cst/ltp_1804289383/tasks' at tst_cgroup.c:296: ENOSPC (28) So these lines will be add in to library as neccesary: --- a/lib/tst_cgroup.c +++ b/lib/tst_cgroup.c @@ -251,6 +251,16 @@ void tst_cgroup_mount(enum tst_cgroup_ctrl ctrl, const char *cgroup_dir) break; case TST_CGROUP_CPUSET: tst_cgroup1_mount("cpusetcg", "cpuset", cgroup_dir, cgroup_new_dir); + /* + * we should assign one or more memory nodes to cpuset.mems + * and cpuset.cpus, otherwise, echo $$ > tasks gives “no space + * left on device: ENOSPC” when trying to use cpuset. + * + * Or, setting cgroup.clone_children to 1 can help in automatically + * inheriting memory and node setting from parent cgroup when a + * child cgroup is created. + */ + tst_cgroup_set_knob(cgroup_dir, "../cgroup.clone_children", 1); break;
I have pushed the patchset with the following changes: 1. use SAFE_MALLOC to replace SAFE_MMAP 2. set 1 to cgroup.clone_children before creating dir 3. add more code comments in tst_cgroup.h Let's see how things going on with these new APIs using in next.
Hi all,
> I have pushed the patchset with the following changes:
I've updated wiki. Please remember updating it when you touch doc/*.txt.
Kind regards,
Petr
On Mon, Jun 22, 2020 at 10:20 PM Petr Vorel <pvorel@suse.cz> wrote: > Hi all, > > > I have pushed the patchset with the following changes: > I've updated wiki. Please remember updating it when you touch doc/*.txt. > Ah yes, I have forgotten that more than twice. Sorry for the inconvenience.
Hello, Li Wang <liwang@redhat.com> writes: > Many of our LTP tests need Control Group in the configuration, > this patch makes cgroup unified mounting at setup phase to be > possible. The method is extracted from mem.h with the purpose > of extendible for further developing, and trying to compatible > the current two versions of cgroup, > > It's hard to make all APIs be strictly consistent because there > are many differences between v1 and v2. But it capsulate the detail > of cgroup mounting in high-level functions, which will be easier > to use cgroup without considering too much technical thing. > > Btw, test get passed on RHEL7(x86_64), RHEL8(ppc64le), > Fedora32(x86_64). This appears to be broken on SUSE. > +enum tst_cgroup_ver tst_cgroup_version(void) > +{ > + if (tst_cgroup_check("cgroup2")) { > + if (!tst_is_mounted("cgroup2") && tst_is_mounted("cgroup")) Unfortunately this is backwards: if a cgroup controller is mounted (e.g. memory) in the v1 separated hierarchy then it is not available in the cgroup v2 unified hierarchy even though the v2 unified hierarchy exists. So ksm{03,04} fails with: safe_file_ops.c:299: BROK: Failed to close FILE '/tmp/cgroup_mem/cgroup.subtree_control' at tst_cgroup.c:301: ENOENT (2) because we have the memory controller (in fact all the controllers) mounted in the V1 way, but we also have the cgroup2 mounted. If I unmount the memory controller from V1 then the test passes. Another potential problem is that it may be possible to remove access to controllers some other way in which case you have to check cgroup.controllers to see what is available. OTOH maybe this is not possible because you mount a new cgroup2 hierarchy?
On Wed, Sep 23, 2020 at 7:03 PM Richard Palethorpe <rpalethorpe@suse.de> wrote: > Hello, > > Li Wang <liwang@redhat.com> writes: > > > Many of our LTP tests need Control Group in the configuration, > > this patch makes cgroup unified mounting at setup phase to be > > possible. The method is extracted from mem.h with the purpose > > of extendible for further developing, and trying to compatible > > the current two versions of cgroup, > > > > It's hard to make all APIs be strictly consistent because there > > are many differences between v1 and v2. But it capsulate the detail > > of cgroup mounting in high-level functions, which will be easier > > to use cgroup without considering too much technical thing. > > > > Btw, test get passed on RHEL7(x86_64), RHEL8(ppc64le), > > Fedora32(x86_64). > > This appears to be broken on SUSE. > > > +enum tst_cgroup_ver tst_cgroup_version(void) > > +{ > > + if (tst_cgroup_check("cgroup2")) { > > + if (!tst_is_mounted("cgroup2") && tst_is_mounted("cgroup")) > > Unfortunately this is backwards: if a cgroup controller is mounted > (e.g. memory) in the v1 separated hierarchy then it is not available in > the cgroup v2 unified hierarchy even though the v2 unified hierarchy > exists. > That's on purpose because we don't want the user to mix v1 and v2 in using. In other words, if a version of cgroup has been chosen on a SUT then LTP follows in the same version. The cgroup lib choose cgroup version likes: * system doesn't support any cgroup(v1, v2): TCONF * system only support cgroup v1: choose v1 * system only support cgroup v2: choose v2 * system support v1 & v2 but mounting v1: chosse v1 <=== this is what you met * system support v1 & v2 but mounting v2: choose v2 * system support v1 & v2 but mounting none: choose v2 > > So ksm{03,04} fails with: > > safe_file_ops.c:299: BROK: Failed to close FILE > '/tmp/cgroup_mem/cgroup.subtree_control' at tst_cgroup.c:301: ENOENT (2) > > because we have the memory controller (in fact all the controllers) > mounted in the V1 way, but we also have the cgroup2 mounted. If I > Hmm, this is a situation that system mount v1&v2 at the same time. So the ksm03/4 choose v2 automatically but hit a problem in cgroup_v2 using. I pesonally think it mainly imputes to the SUT config in cgroup, because that makes cgroup_v2 has no actuall controllers in the cgroup.controllers. After umounting all V1 hierarchies, the controllers get back into V2. > unmount the memory controller from V1 then the test passes. > > Another potential problem is that it may be possible to remove access to > controllers some other way in which case you have to check > cgroup.controllers to see what is available. OTOH maybe this is not > possible because you mount a new cgroup2 hierarchy? > I'm not sure about this. To be honest, I hope a Linux distribution provides a pure way in cgroup using, which means just mounting V1 or V2 but not to mix together. Or, do you think LTP should handle the situation for the v1&v2 mixing mount?
Hello Li, Li Wang <liwang@redhat.com> writes: > On Wed, Sep 23, 2020 at 7:03 PM Richard Palethorpe <rpalethorpe@suse.de> > wrote: > >> Hello, >> >> Li Wang <liwang@redhat.com> writes: >> >> > Many of our LTP tests need Control Group in the configuration, >> > this patch makes cgroup unified mounting at setup phase to be >> > possible. The method is extracted from mem.h with the purpose >> > of extendible for further developing, and trying to compatible >> > the current two versions of cgroup, >> > >> > It's hard to make all APIs be strictly consistent because there >> > are many differences between v1 and v2. But it capsulate the detail >> > of cgroup mounting in high-level functions, which will be easier >> > to use cgroup without considering too much technical thing. >> > >> > Btw, test get passed on RHEL7(x86_64), RHEL8(ppc64le), >> > Fedora32(x86_64). >> >> This appears to be broken on SUSE. >> >> > +enum tst_cgroup_ver tst_cgroup_version(void) >> > +{ >> > + if (tst_cgroup_check("cgroup2")) { >> > + if (!tst_is_mounted("cgroup2") && tst_is_mounted("cgroup")) >> >> Unfortunately this is backwards: if a cgroup controller is mounted >> (e.g. memory) in the v1 separated hierarchy then it is not available in >> the cgroup v2 unified hierarchy even though the v2 unified hierarchy >> exists. >> > > That's on purpose because we don't want the user to mix v1 and v2 in using. > In other words, if a version of cgroup has been chosen on a SUT then LTP > follows in the same version. > > The cgroup lib choose cgroup version likes: > * system doesn't support any cgroup(v1, v2): TCONF > * system only support cgroup v1: choose v1 > * system only support cgroup v2: choose v2 > * system support v1 & v2 but mounting v1: chosse v1 <=== this is > what you met Perhaps you meant: if (tst_is_mounted("cgroup")) cg_ver = TST_CGROUP_V1; else cg_ver = TST_CGROUP_V2; There is no point trying to use V2 if the controller is active in V1. > * system support v1 & v2 but mounting v2: choose v2 > * system support v1 & v2 but mounting none: choose v2 > > > >> >> So ksm{03,04} fails with: >> >> safe_file_ops.c:299: BROK: Failed to close FILE >> '/tmp/cgroup_mem/cgroup.subtree_control' at tst_cgroup.c:301: ENOENT (2) >> >> because we have the memory controller (in fact all the controllers) >> mounted in the V1 way, but we also have the cgroup2 mounted. If I >> > > Hmm, this is a situation that system mount v1&v2 at the same time. > > So the ksm03/4 choose v2 automatically but hit a problem in cgroup_v2 using. > I pesonally think it mainly imputes to the SUT config in cgroup, because > that > makes cgroup_v2 has no actuall controllers in the cgroup.controllers. After > umounting all V1 hierarchies, the controllers get back into V2. > yes > > >> unmount the memory controller from V1 then the test passes. >> >> Another potential problem is that it may be possible to remove access to >> controllers some other way in which case you have to check >> cgroup.controllers to see what is available. OTOH maybe this is not >> possible because you mount a new cgroup2 hierarchy? >> > > I'm not sure about this. > > To be honest, I hope a Linux distribution provides a pure way in cgroup > using, which means just mounting V1 or V2 but not to mix together. Or, do > you think LTP should handle the situation for the v1&v2 mixing mount? For now, I think we just need to use V1 if it is mounted. I don't like the fact we have both mounted, but it seems most software can handle it, so LTP should as well.
On Wed, Sep 23, 2020 at 9:14 PM Richard Palethorpe <rpalethorpe@suse.de> wrote: > ... > > That's on purpose because we don't want the user to mix v1 and v2 in > using. > > In other words, if a version of cgroup has been chosen on a SUT then LTP > > follows in the same version. > > > > The cgroup lib choose cgroup version likes: > > * system doesn't support any cgroup(v1, v2): TCONF > > * system only support cgroup v1: choose v1 > > * system only support cgroup v2: choose v2 > > * system support v1 & v2 but mounting v1: chosse v1 <=== this is > > what you met > > Perhaps you meant: > > if (tst_is_mounted("cgroup")) > cg_ver = TST_CGROUP_V1; > else > cg_ver = TST_CGROUP_V2; > This is not enough for the version choice. As tst_is_mounted(“cgroup”) can't distinguish precisely what "cgroup" or "cgroup2" has been mounted. Because the function is fuzzy matching a string. > > > > To be honest, I hope a Linux distribution provides a pure way in cgroup > > using, which means just mounting V1 or V2 but not to mix together. Or, do > > you think LTP should handle the situation for the v1&v2 mixing mount? > > For now, I think we just need to use V1 if it is mounted. I don't like > the fact we have both mounted, but it seems most software can handle it, > so LTP should as well. > Agreed.
Hello, Li Wang <liwang@redhat.com> writes: >> Perhaps you meant: >> >> if (tst_is_mounted("cgroup")) >> cg_ver = TST_CGROUP_V1; >> else >> cg_ver = TST_CGROUP_V2; >> > > This is not enough for the version choice. As tst_is_mounted(“cgroup”) can't > distinguish precisely what "cgroup" or "cgroup2" has been mounted. Because > the function is fuzzy matching a string. >> > >> > To be honest, I hope a Linux distribution provides a pure way in cgroup >> > using, which means just mounting V1 or V2 but not to mix together. Or, do >> > you think LTP should handle the situation for the v1&v2 mixing mount? >> >> For now, I think we just need to use V1 if it is mounted. I don't like >> the fact we have both mounted, but it seems most software can handle it, >> so LTP should as well. >> > > Agreed. OK, I will send a patch unless you send one first.
diff --git a/doc/test-writing-guidelines.txt b/doc/test-writing-guidelines.txt index 93ca506d9..42b360da0 100644 --- a/doc/test-writing-guidelines.txt +++ b/doc/test-writing-guidelines.txt @@ -2053,6 +2053,50 @@ the prefix field of file pointed by path equals to the value passed to this func Also having a similar api pair TST_ASSERT_FILE_INT/STR(path, prefix, val) to assert the field value of file. +2.2.36 Using Control Group +^^^^^^^^^^^^^^^^^^^^^^^^^^ +Some of LTP tests need Control Group in their configuration, tst_cgroup.h provides +APIs to make cgroup unified mounting at setup phase to be possible. The method is +extracted from mem.h with the purpose of extendible for further developing, and +trying to compatible the current two versions of cgroup. + +Considering there are many differences between cgroup v1 and v2, here we capsulate +the detail of cgroup mounting in high-level functions, which will be easier to use +cgroup without caring about too much technical thing. + +Also, we do cgroup mount/umount work for the different hierarchy automatically. + +[source,c] +------------------------------------------------------------------------------- +#include "tst_test.h" + +static void run(void) +{ + ... + + tst_cgroup_move_current(PATH_TMP_CG_MEM); + tst_cgroup_mem_set_maxbytes(PATH_TMP_CG_MEM, MEMSIZE); + + // do test under cgroup + ... +} + +static void setup(void) +{ + tst_cgroup_mount(TST_CGROUP_MEMCG, PATH_TMP_CG_MEM); +} + +static void cleanup(void) +{ + tst_cgroup_umount(PATH_TMP_CG_MEM); +} + +struct tst_test test = { + .test_all = run, + ... +}; + +------------------------------------------------------------------------------- 2.3 Writing a testcase in shell ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ diff --git a/include/tst_cgroup.h b/include/tst_cgroup.h new file mode 100644 index 000000000..0bc73b542 --- /dev/null +++ b/include/tst_cgroup.h @@ -0,0 +1,39 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright (c) 2020 Red Hat, Inc. + * Copyright (c) 2020 Li Wang <liwang@redhat.com> + */ + +#ifndef TST_CGROUP_H +#define TST_CGROUP_H + +#define PATH_TMP_CG_MEM "/tmp/cgroup_mem" +#define PATH_TMP_CG_CST "/tmp/cgroup_cst" + +enum tst_cgroup_ver { + TST_CGROUP_V1 = 1, + TST_CGROUP_V2 = 2, +}; + +enum tst_cgroup_ctrl { + TST_CGROUP_MEMCG = 1, + TST_CGROUP_CPUSET = 2, + /* add cgroup controller */ +}; + +enum tst_cgroup_ver tst_cgroup_version(void); + +void tst_cgroup_mount(enum tst_cgroup_ctrl ctrl, const char *cgroup_dir); +void tst_cgroup_umount(const char *cgroup_dir); + +void tst_cgroup_move_current(const char *cgroup_dir); +void tst_cgroup_set_knob(const char *cgroup_dir, const char *knob, long value); + +void tst_cgroup_mem_set_maxbytes(const char *cgroup_dir, long memsz); +int tst_cgroup_mem_swapacct_enabled(const char *cgroup_dir); +void tst_cgroup_mem_set_maxswap(const char *cgroup_dir, long memsz); + +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); + +#endif /* TST_CGROUP_H */ diff --git a/include/tst_test.h b/include/tst_test.h index 5bedb4f6f..b84f7b9dd 100644 --- a/include/tst_test.h +++ b/include/tst_test.h @@ -39,6 +39,7 @@ #include "tst_capability.h" #include "tst_hugepage.h" #include "tst_assert.h" +#include "tst_cgroup.h" /* * Reports testcase result. diff --git a/lib/newlib_tests/test21.c b/lib/newlib_tests/test21.c new file mode 100644 index 000000000..c3fa5584e --- /dev/null +++ b/lib/newlib_tests/test21.c @@ -0,0 +1,76 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright (c) 2020 Red Hat, Inc. + * Copyright (c) 2020 Li Wang <liwang@redhat.com> + */ + +/* + * Tests tst_cgroup.h APIs + */ + +#include "tst_test.h" +#include "tst_cgroup.h" + +#define PATH_CGROUP1 "/mnt/liwang1" +#define PATH_CGROUP2 "/mnt/liwang2" +#define MEMSIZE 1024 * 1024 + +static void do_test(void) +{ + /* we should assign one or more memory nodes to cpuset.mems + * and cpuset.cpus, otherwise, echo $$ > tasks gives “no space + * left on device” when trying to use cpuset. + * + * Or, setting cgroup.clone_children to 1 can help in automatically + * inheriting memory and node setting from parent cgroup when + * a child cgroup is created. + */ + tst_cgroup_set_knob(PATH_TMP_CG_CST, "../cgroup.clone_children", 1); + + pid_t pid = SAFE_FORK(); + + switch (pid) { + case 0: + tst_cgroup_move_current(PATH_CGROUP1); + tst_cgroup_mem_set_maxbytes(PATH_CGROUP1, MEMSIZE); + tst_cgroup_mem_set_maxswap(PATH_CGROUP1, MEMSIZE); + + tst_cgroup_move_current(PATH_CGROUP2); + + break; + default: + tst_cgroup_move_current(PATH_TMP_CG_CST); + + tst_cgroup_move_current(PATH_TMP_CG_MEM); + tst_cgroup_mem_set_maxbytes(PATH_TMP_CG_MEM, MEMSIZE); + tst_cgroup_mem_set_maxswap(PATH_TMP_CG_MEM, MEMSIZE); + break; + } + + tst_res(TPASS, "Cgroup mount test"); +} + +static void setup(void) +{ + tst_cgroup_mount(TST_CGROUP_MEMCG, PATH_TMP_CG_MEM); + tst_cgroup_mount(TST_CGROUP_MEMCG, PATH_CGROUP1); + + tst_cgroup_mount(TST_CGROUP_CPUSET, PATH_TMP_CG_CST); + tst_cgroup_mount(TST_CGROUP_CPUSET, PATH_CGROUP2); +} + +static void cleanup(void) +{ + tst_cgroup_umount(PATH_TMP_CG_MEM); + tst_cgroup_umount(PATH_CGROUP1); + + tst_cgroup_umount(PATH_TMP_CG_CST); + tst_cgroup_umount(PATH_CGROUP2); +} + +static struct tst_test test = { + .test_all = do_test, + .setup = setup, + .cleanup = cleanup, + .forks_child = 1, +}; diff --git a/lib/tst_cgroup.c b/lib/tst_cgroup.c new file mode 100644 index 000000000..f918e7e85 --- /dev/null +++ b/lib/tst_cgroup.c @@ -0,0 +1,420 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright (c) 2020 Red Hat, Inc. + * Copyright (c) 2020 Li Wang <liwang@redhat.com> + */ + +#define TST_NO_DEFAULT_MAIN + +#include <stdio.h> +#include <stdlib.h> +#include <sys/mount.h> + +#include "tst_test.h" +#include "tst_safe_macros.h" +#include "tst_safe_stdio.h" +#include "tst_cgroup.h" +#include "tst_device.h" + +static enum tst_cgroup_ver tst_cg_ver; + +static int tst_cgroup_check(const char *cgroup) +{ + char line[PATH_MAX]; + FILE *file; + int cg_check = 0; + + file = SAFE_FOPEN("/proc/filesystems", "r"); + while (fgets(line, sizeof(line), file)) { + if (strstr(line, cgroup) != NULL) { + cg_check = 1; + break; + } + } + SAFE_FCLOSE(file); + + return cg_check; +} + +enum tst_cgroup_ver tst_cgroup_version(void) +{ + if (tst_cgroup_check("cgroup2")) { + if (!tst_is_mounted("cgroup2") && tst_is_mounted("cgroup")) + return TST_CGROUP_V1; + else + return TST_CGROUP_V2; + } + + if (tst_cgroup_check("cgroup")) + return TST_CGROUP_V1; + + tst_brk(TCONF, "Cgroup is not configured"); +} + +static void tst_cgroup1_mount(const char *name, const char *option, + const char *mnt_path, const char *new_path) +{ + if (tst_is_mounted(mnt_path)) + goto out; + + SAFE_MKDIR(mnt_path, 0777); + if (mount(name, mnt_path, "cgroup", 0, option) == -1) { + if (errno == ENODEV) { + if (rmdir(mnt_path) == -1) + tst_res(TWARN | TERRNO, "rmdir %s failed", mnt_path); + tst_brk(TCONF, + "Cgroup v1 is not configured in kernel"); + } + tst_brk(TBROK | TERRNO, "mount %s", mnt_path); + } + +out: + SAFE_MKDIR(new_path, 0777); + + tst_res(TINFO, "Cgroup(%s) v1 mount at %s success", option, mnt_path); +} + +static void tst_cgroup2_mount(const char *mnt_path, const char *new_path) +{ + if (tst_is_mounted(mnt_path)) + goto out; + + SAFE_MKDIR(mnt_path, 0777); + if (mount("cgroup2", mnt_path, "cgroup2", 0, NULL) == -1) { + if (errno == ENODEV) { + if (rmdir(mnt_path) == -1) + tst_res(TWARN | TERRNO, "rmdir %s failed", mnt_path); + tst_brk(TCONF, + "Cgroup v2 is not configured in kernel"); + } + tst_brk(TBROK | TERRNO, "mount %s", mnt_path); + } + +out: + SAFE_MKDIR(new_path, 0777); + + tst_res(TINFO, "Cgroup v2 mount at %s success", mnt_path); +} + +static void tst_cgroupN_umount(const char *mnt_path, const char *new_path) +{ + FILE *fp; + int fd; + char s_new[BUFSIZ], s[BUFSIZ], value[BUFSIZ]; + + if (!tst_is_mounted(mnt_path)) + return; + + /* Move all processes in task(v2: cgroup.procs) to its parent node. */ + if (tst_cg_ver & TST_CGROUP_V1) + sprintf(s, "%s/tasks", mnt_path); + if (tst_cg_ver & TST_CGROUP_V2) + sprintf(s, "%s/cgroup.procs", mnt_path); + + fd = open(s, O_WRONLY); + if (fd == -1) + tst_res(TWARN | TERRNO, "open %s", s); + + if (tst_cg_ver & TST_CGROUP_V1) + snprintf(s_new, BUFSIZ, "%s/tasks", new_path); + if (tst_cg_ver & TST_CGROUP_V2) + snprintf(s_new, BUFSIZ, "%s/cgroup.procs", new_path); + + fp = fopen(s_new, "r"); + if (fp == NULL) + tst_res(TWARN | TERRNO, "fopen %s", s_new); + if ((fd != -1) && (fp != NULL)) { + while (fgets(value, BUFSIZ, fp) != NULL) + if (write(fd, value, strlen(value) - 1) + != (ssize_t)strlen(value) - 1) + tst_res(TWARN | TERRNO, "write %s", s); + } + if (fd != -1) + close(fd); + if (fp != NULL) + fclose(fp); + if (rmdir(new_path) == -1) + tst_res(TWARN | TERRNO, "rmdir %s", new_path); + if (umount(mnt_path) == -1) + tst_res(TWARN | TERRNO, "umount %s", mnt_path); + if (rmdir(mnt_path) == -1) + tst_res(TWARN | TERRNO, "rmdir %s", mnt_path); + + if (tst_cg_ver & TST_CGROUP_V1) + tst_res(TINFO, "Cgroup v1 unmount success"); + if (tst_cg_ver & TST_CGROUP_V2) + tst_res(TINFO, "Cgroup v2 unmount success"); +} + +struct tst_cgroup_path { + char *mnt_path; + char *new_path; + struct tst_cgroup_path *next; +}; + +static struct tst_cgroup_path *tst_cgroup_paths; + +static void tst_cgroup_set_path(const char *cgroup_dir) +{ + char cgroup_new_dir[PATH_MAX]; + struct tst_cgroup_path *tst_cgroup_path, *a; + + if (!cgroup_dir) + tst_brk(TBROK, "Invalid cgroup dir, plese check cgroup_dir"); + + sprintf(cgroup_new_dir, "%s/ltp_%d", cgroup_dir, rand()); + + /* To store cgroup path in the 'path' list */ + tst_cgroup_path = SAFE_MMAP(NULL, (sizeof(struct tst_cgroup_path)), + PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + tst_cgroup_path->mnt_path = SAFE_MMAP(NULL, (strlen(cgroup_dir) + 1), + PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + tst_cgroup_path->new_path = SAFE_MMAP(NULL, (strlen(cgroup_new_dir) + 1), + PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + + if (!tst_cgroup_paths) { + tst_cgroup_paths = tst_cgroup_path; + } else { + a = tst_cgroup_paths; + do { + if (!a->next) { + a->next = tst_cgroup_path; + break; + } + a = a->next; + } while (a); + } + + sprintf(tst_cgroup_path->mnt_path, "%s", cgroup_dir); + sprintf(tst_cgroup_path->new_path, "%s", cgroup_new_dir); +} + +static char *tst_cgroup_get_path(const char *cgroup_dir) +{ + struct tst_cgroup_path *a; + + if (!tst_cgroup_paths) + return NULL; + + a = tst_cgroup_paths; + + while (strcmp(a->mnt_path, cgroup_dir) != 0){ + if (!a->next) { + tst_res(TINFO, "%s is not found", cgroup_dir); + return NULL; + } + a = a->next; + }; + + return a->new_path; +} + +static void tst_cgroup_del_path(const char *cgroup_dir) +{ + struct tst_cgroup_path *a, *b; + + if (tst_cgroup_paths) + return; + + a = b = tst_cgroup_paths; + + while (strcmp(b->mnt_path, cgroup_dir) != 0) { + if (!b->next) { + tst_res(TINFO, "%s is not found", cgroup_dir); + return; + } + a = b; + b = b->next; + }; + + if (b == tst_cgroup_paths) + tst_cgroup_paths = b->next; + else + a->next = b->next; + + SAFE_MUNMAP(b->mnt_path, sizeof(b->mnt_path)); + SAFE_MUNMAP(b->new_path, sizeof(b->new_path)); + SAFE_MUNMAP(b, sizeof(b)); +} + +void tst_cgroup_mount(enum tst_cgroup_ctrl ctrl, const char *cgroup_dir) +{ + char *cgroup_new_dir; + char knob_path[PATH_MAX]; + + tst_cg_ver = tst_cgroup_version(); + + tst_cgroup_set_path(cgroup_dir); + cgroup_new_dir = tst_cgroup_get_path(cgroup_dir); + + if (tst_cg_ver & TST_CGROUP_V1) { + switch(ctrl) { + case TST_CGROUP_MEMCG: + tst_cgroup1_mount("memcg", "memory", cgroup_dir, cgroup_new_dir); + break; + case TST_CGROUP_CPUSET: + tst_cgroup1_mount("cpusetcg", "cpuset", cgroup_dir, cgroup_new_dir); + break; + default: + tst_brk(TBROK, "Invalid cgroup controller: %d", ctrl); + } + } + + if (tst_cg_ver & TST_CGROUP_V2) { + tst_cgroup2_mount(cgroup_dir, cgroup_new_dir); + + switch(ctrl) { + case TST_CGROUP_MEMCG: + sprintf(knob_path, "%s/cgroup.subtree_control", cgroup_dir); + SAFE_FILE_PRINTF(knob_path, "%s", "+memory"); + break; + case TST_CGROUP_CPUSET: + tst_brk(TCONF, "Cgroup v2 hasn't achieve cpuset subsystem"); + break; + default: + tst_brk(TBROK, "Invalid cgroup controller: %d", ctrl); + } + } +} + +void tst_cgroup_umount(const char *cgroup_dir) +{ + char *cgroup_new_dir; + + cgroup_new_dir = tst_cgroup_get_path(cgroup_dir); + tst_cgroupN_umount(cgroup_dir, cgroup_new_dir); + tst_cgroup_del_path(cgroup_dir); +} + +void tst_cgroup_set_knob(const char *cgroup_dir, const char *knob, long value) +{ + char *cgroup_new_dir; + char knob_path[PATH_MAX]; + + cgroup_new_dir = tst_cgroup_get_path(cgroup_dir); + sprintf(knob_path, "%s/%s", cgroup_new_dir, knob); + SAFE_FILE_PRINTF(knob_path, "%ld", value); +} + +void tst_cgroup_move_current(const char *cgroup_dir) +{ + if (tst_cg_ver & TST_CGROUP_V1) + tst_cgroup_set_knob(cgroup_dir, "tasks", getpid()); + + if (tst_cg_ver & TST_CGROUP_V2) + tst_cgroup_set_knob(cgroup_dir, "cgroup.procs", getpid()); +} + +void tst_cgroup_mem_set_maxbytes(const char *cgroup_dir, long memsz) +{ + if (tst_cg_ver & TST_CGROUP_V1) + tst_cgroup_set_knob(cgroup_dir, "memory.limit_in_bytes", memsz); + + if (tst_cg_ver & TST_CGROUP_V2) + tst_cgroup_set_knob(cgroup_dir, "memory.max", memsz); +} + +int tst_cgroup_mem_swapacct_enabled(const char *cgroup_dir) +{ + char *cgroup_new_dir; + char knob_path[PATH_MAX]; + + cgroup_new_dir = tst_cgroup_get_path(cgroup_dir); + + if (tst_cg_ver & TST_CGROUP_V1) { + sprintf(knob_path, "%s/%s", + cgroup_new_dir, "/memory.memsw.limit_in_bytes"); + + if ((access(knob_path, F_OK) == -1)) { + if (errno == ENOENT) + tst_res(TCONF, "memcg swap accounting is disabled"); + else + tst_brk(TBROK | TERRNO, "failed to access %s", knob_path); + } else { + return 1; + } + } + + if (tst_cg_ver & TST_CGROUP_V2) { + sprintf(knob_path, "%s/%s", + cgroup_new_dir, "/memory.swap.max"); + + if ((access(knob_path, F_OK) == -1)) { + if (errno == ENOENT) + tst_res(TCONF, "memcg swap accounting is disabled"); + else + tst_brk(TBROK | TERRNO, "failed to access %s", knob_path); + } else { + return 1; + } + } + + return 0; +} + +void tst_cgroup_mem_set_maxswap(const char *cgroup_dir, long memsz) +{ + if (tst_cg_ver & TST_CGROUP_V1) + tst_cgroup_set_knob(cgroup_dir, "memory.memsw.limit_in_bytes", memsz); + + if (tst_cg_ver & TST_CGROUP_V2) + 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) +{ + int fd; + char *cgroup_new_dir; + char knob_path[PATH_MAX]; + + cgroup_new_dir = tst_cgroup_get_path(cgroup_dir); + + /* + * try either '/dev/cpuset/XXXX' or '/dev/cpuset/cpuset.XXXX' + * please see Documentation/cgroups/cpusets.txt from kernel src + * for details + */ + sprintf(knob_path, "%s/%s", cgroup_new_dir, filename); + fd = open(knob_path, O_RDONLY); + if (fd == -1) { + if (errno == ENOENT) { + sprintf(knob_path, "%s/cpuset.%s", + cgroup_new_dir, filename); + fd = SAFE_OPEN(knob_path, O_RDONLY); + } else + tst_brk(TBROK | TERRNO, "open %s", knob_path); + } + + if (read(fd, retbuf, sizeof(retbuf)) < 0) + tst_brk(TBROK | TERRNO, "read %s", knob_path); + + close(fd); +} + +void tst_cgroup_cpuset_write_files(const char *cgroup_dir, const char *filename, const char *buf) +{ + int fd; + char *cgroup_new_dir; + char knob_path[PATH_MAX]; + + cgroup_new_dir = tst_cgroup_get_path(cgroup_dir); + + /* + * try either '/dev/cpuset/XXXX' or '/dev/cpuset/cpuset.XXXX' + * please see Documentation/cgroups/cpusets.txt from kernel src + * for details + */ + sprintf(knob_path, "%s/%s", cgroup_new_dir, filename); + fd = open(knob_path, O_WRONLY); + if (fd == -1) { + if (errno == ENOENT) { + sprintf(knob_path, "%s/cpuset.%s", cgroup_new_dir, filename); + fd = SAFE_OPEN(knob_path, O_WRONLY); + } else + tst_brk(TBROK | TERRNO, "open %s", knob_path); + } + + SAFE_WRITE(1, fd, buf, strlen(buf)); + + close(fd); +}
Many of our LTP tests need Control Group in the configuration, this patch makes cgroup unified mounting at setup phase to be possible. The method is extracted from mem.h with the purpose of extendible for further developing, and trying to compatible the current two versions of cgroup, It's hard to make all APIs be strictly consistent because there are many differences between v1 and v2. But it capsulate the detail of cgroup mounting in high-level functions, which will be easier to use cgroup without considering too much technical thing. Btw, test get passed on RHEL7(x86_64), RHEL8(ppc64le), Fedora32(x86_64). Signed-off-by: Li Wang <liwang@redhat.com> --- Notes: v3 --> v4 * remove tst_cgroup_mnt/new_path * tst_cgroup_get_path() return a pointer to new_path * add tst_cgroup_cpuset_read/write_files() doc/test-writing-guidelines.txt | 44 ++++ include/tst_cgroup.h | 39 +++ include/tst_test.h | 1 + lib/newlib_tests/test21.c | 76 ++++++ lib/tst_cgroup.c | 420 ++++++++++++++++++++++++++++++++ 5 files changed, 580 insertions(+) create mode 100644 include/tst_cgroup.h create mode 100644 lib/newlib_tests/test21.c create mode 100644 lib/tst_cgroup.c