Extend idmapped mount testsuite
v4 cover Extend idmapped mount testsuite
From: Christian Brauner <[email protected]>
Hey everyone,
This is v4 with only a minimal change to getopt_long_only(). The rest is identical. In v3 the changes requested by Christoph and his Reviewed-bys added. I've also added Josef's Reviewed-by for the newly added btrfs specific idmapped mount testsuite. There are no major changes to v2.
This time around I've put everyone in To: to make sure that they receive all patches. The list apparently still refues patches if they are fairly huge. The series can be pulled from three locations:
[email protected]:pub/scm/linux/kernel/git/brauner/xfstests-dev fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns
This adds three new tests:
- a regression test for vfs capabilities
- a new test with nested and complex idmapping layouts
- a new btrfs specific idmapped mount testsuite
Since v2 patches to support idmapped mounts on btrfs have been merged into btrfs' for-next tree:
https://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux.git/commit/?h=for-next&id=f691ae1e0bc591df5e9897387d726adbb440fc4c
and so are on track to be included for v5.15. I'd like to get the btrfs tests merged now so linux-next can be tested with the them. I hope that's ok.
I ran:
git rebase -i -x "make -j4 clean && make -j4 && sudo ./check -g idmapped" HEAD~8 on the whole series. Everthing passes.
Thanks!
Christian
Christian Brauner (8):
idmapped-mounts: use die() helper
idmapped-mounts: switch to getopt_long_only()
idmapped-mounts: introduce an explicit command line switch for
testsuite
generic/640: add fscaps regression test
idmapped-mounts: refactor helpers
idmapped-mounts: add nested userns creation helpers generic/641: add nested user namespace tests btrfs/244: introduce btrfs specific idmapped mounts tests
configure.ac | 10 +- src/idmapped-mounts/idmapped-mounts.c | 4747 ++++++++++++++++++++++++- src/idmapped-mounts/mount-idmapped.c | 229 +- src/idmapped-mounts/utils.c | 359 +- src/idmapped-mounts/utils.h | 102 +- tests/btrfs/244 | 34 + tests/btrfs/244.out | 2 + tests/generic/633 | 3 +- tests/generic/640 | 28 + tests/generic/640.out | 2 + tests/generic/641 | 28 + tests/generic/641.out | 2 + 12 files changed, 5231 insertions(+), 315 deletions(-) create mode 100755 tests/btrfs/244 create mode 100644 tests/btrfs/244.out create mode 100755 tests/generic/640 create mode 100644 tests/generic/640.out create mode 100755 tests/generic/641 create mode 100644 tests/generic/641.out
base-commit: dad0c0a852d1b10e7da285f29e99397dec0efec1
Signature · 1 line
2.30.2
PATCH v4 1/8 idmapped-mounts: use die() helper
From: Christian Brauner <[email protected]>
Use the dedicated helper to report an error and exit with failure instead of hand-rolling it.
---
/* v2 */
patch not present
/* v3 */
- Christoph Hellwig <[email protected]>:
- Split into separate patch.
/* v4 */
unchanged
--- src/idmapped-mounts/idmapped-mounts.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)
src/idmapped-mounts/idmapped-mounts.c +2 -4
@@ -8804,10 +8804,8 @@ static bool run_test(struct t_idmapped_mounts suite[], size_t suite_size)if (pid ==0 ) { ret = t->test (); -if (ret) { -fprintf (stderr,"failure: %s\n" , t->description); -exit (EXIT_FAILURE ); - } +if (ret) +die ("failure: %s" , t->description);exit (EXIT_SUCCESS ); }
Signature · 1 line
2.30.2
PATCH v4 2/8 idmapped-mounts: switch to getopt_long_only()
In reply to Christian BraunerFrom: Christian Brauner <[email protected]>
We're not using the shortopts anywhere anyway and shouldn't encourage using shortopts. It's much more descriptive to see:
$here/src/idmapped-mounts/idmapped-mounts \
--test-btrfs \
--device "$TEST_DEV" \
--mountpoint "$TEST_DIR" \
--scratch-device "$SCRATCH_DEV" \
--scratch-mountpoint "$SCRATCH_MNT"
--fstype "$FSTYP"in a test than it is to see:
$here/src/idmapped-mounts/idmapped-mounts \
-b
-d "$TEST_DEV" \
-m "$TEST_DIR" \
-s "$SCRATCH_DEV" \
-a "$SCRATCH_MNT" \
-f "$FSTYP"In the second case one has to go consult the source code to make sure that the correct option is passed. In the first case one can just see it directly.
---
/* v2 */
patch not present
/* v3 */
- Christoph Hellwig <[email protected]>:
- Split into separate patch.
/* v4 */
- Christoph Hellwig <[email protected]>:
- Pass empty string for optstring argument in getopt_long_only().
--- src/idmapped-mounts/idmapped-mounts.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
src/idmapped-mounts/idmapped-mounts.c +6 -3
@@ -8717,8 +8717,11 @@ static void usage(void)fprintf (stderr," Run idmapped mount tests\n\n" );fprintf (stderr,"Arguments:\n" ); -fprintf (stderr,"-d --device Device used in the tests\n" ); -fprintf (stderr,"-m --mountpoint Mountpoint of device\n" ); +fprintf (stderr,"--device Device used in the tests\n" ); +fprintf (stderr,"--fstype Filesystem type used in the tests\n" ); +fprintf (stderr,"--help Print help\n" ); +fprintf (stderr,"--mountpoint Mountpoint of device\n" ); +fprintf (stderr,"--supported Test whether idmapped mounts are supported on this filesystem\n" );_exit (EXIT_SUCCESS ); }
@@ -8826,7 +8829,7 @@ int main(int argc, char *argv[])int index =0 ;bool supported = false; -while ((ret =getopt_long (argc, argv,"" , longopts, &index)) !=-1 ) { +while ((ret =getopt_long_only (argc, argv,"" , longopts, &index)) !=-1 ) {switch (ret) {case 'd' : t_device = optarg;
Signature · 1 line
2.30.2
PATCH v4 3/8 idmapped-mounts: introduce an explicit command line switch for testsuite
In reply to Christian BraunerFrom: Christian Brauner <[email protected]>
Introduce an explicit command line switch to runs the basic test suite. This prepares for the introduction of additional command line switches to run additional tests.
---
/* v2 */
patch not present
/* v3 */
- Christoph Hellwig <[email protected]>:
- Split into separate patch.
/* v4 */
unchanged
--- src/idmapped-mounts/idmapped-mounts.c | 11 ++++++++--- tests/generic/633 | 3 ++- 2 files changed, 10 insertions(+), 4 deletions(-)
src/idmapped-mounts/idmapped-mounts.c +8 -3
@@ -8722,6 +8722,7 @@ static void usage(void)fprintf (stderr,"--help Print help\n" );fprintf (stderr,"--mountpoint Mountpoint of device\n" );fprintf (stderr,"--supported Test whether idmapped mounts are supported on this filesystem\n" ); +fprintf (stderr,"--test-core Run core idmapped mount testsuite\n" );_exit (EXIT_SUCCESS ); }
@@ -8732,7 +8733,8 @@ static const struct option longopts[] = { {"mountpoint", required_argument, 0, 'm'}, {"supported", no_argument, 0, 's'}, {"help", no_argument, 0, 'h'}, - {NULL, 0, 0, 0 }, + {"test-core", no_argument, 0, 'c'}, + {NULL, 0, 0, 0}, }; struct t_idmapped_mounts {
@@ -8827,7 +8829,7 @@ int main(int argc, char *argv[]) {int fret, ret;int index =0 ; -bool supported = false; +bool supported = false, test_core = false;while ((ret =getopt_long_only (argc, argv,"" , longopts, &index)) !=-1 ) {switch (ret) {
@@ -8843,6 +8845,9 @@ int main(int argc, char *argv[])case 's' : supported = true;break ; +case 'c' : + test_core = true; +break ;case 'h' :/* fallthrough */ default :
@@ -8912,7 +8917,7 @@ int main(int argc, char *argv[]) fret =EXIT_FAILURE ; -if (!run_test (basic_suite,ARRAY_SIZE (basic_suite))) +if (test_core && !run_test (basic_suite,ARRAY_SIZE (basic_suite))) goto out; fret =EXIT_SUCCESS ;
tests/generic/633 +2 -1
@@ -20,7 +20,8 @@ _require_test echo "Silence is golden" -$here/src/idmapped-mounts/idmapped-mounts --device "$TEST_DEV" --mount "$TEST_DIR" --fstype "$FSTYP" +$here/src/idmapped-mounts/idmapped-mounts --test-core --device "$TEST_DEV" \ + --mount "$TEST_DIR" --fstype "$FSTYP" status=$? exit
Signature · 1 line
2.30.2
PATCH v4 4/8 generic/640: add fscaps regression test
In reply to Christian BraunerFrom: Christian Brauner <[email protected]>
Add a test to verify that setting a v3 fscap from an idmapped mount works as expected. This and other related use-cases were regressed by commit [1] which was reverted in [2] and the proper fix merged right before v5.12 was released in [3].
[1]: commit 3b0c2d3eaa83 ("Revert 95ebabde382c ("capabilities: Don't allow writing ambiguous v3 file capabilities")") [2]: commit 95ebabde382c ("capabilities: Don't allow writing ambiguous v3 file capabilities")
[3]: commit db2e718a4798 ("capabilities: require CAP_SETFCAP to map uid 0")
---
/* v2 */
- Eryu Guan <[email protected]>:
- Don't create symlinks for each test. Instead, use the same binary and introduce new options to specify which tests to run.
/* v3 */
unchanged
/* v4 */
unchanged
--- src/idmapped-mounts/idmapped-mounts.c | 159 +++++++++++++++++++++++--- tests/generic/640 | 28 +++++ tests/generic/640.out | 2 + 3 files changed, 174 insertions(+), 15 deletions(-) create mode 100755 tests/generic/640 create mode 100644 tests/generic/640.out
src/idmapped-mounts/idmapped-mounts.c +144 -15
@@ -3199,6 +3199,121 @@ out:return fret; } +static int fscaps_idmapped_mounts_in_userns_valid_in_ancestor_userns (void ) +{ +int fret =-1 ; +int file1_fd = -EBADF , file1_fd2 = -EBADF , open_tree_fd = -EBADF ; +struct mount_attr attr = { + .attr_set =MOUNT_ATTR_IDMAP , + }; +pid_t pid; + + file1_fd =openat (t_dir1_fd,FILE1 ,O_CREAT |O_EXCL |O_CLOEXEC ,0644 ); +if (file1_fd <0 ) { +log_stderr ("failure: openat" ); + goto out; + } + +/* Skip if vfs caps are unsupported. */ +if (set_dummy_vfs_caps (file1_fd,0 ,1000 )) +return 0 ; + +if (fremovexattr (file1_fd,"security.capability" )) { +log_stderr ("failure: fremovexattr" ); + goto out; + } +if (expected_dummy_vfs_caps_uid (file1_fd,-1 )) { +log_stderr ("failure: expected_dummy_vfs_caps_uid" ); + goto out; + } +if (errno !=ENODATA ) { +log_stderr ("failure: errno" ); + goto out; + } + +/* Changing mount properties on a detached mount. */ + attr.userns_fd =get_userns_fd (0 ,10000 ,10000 ); +if (attr.userns_fd <0 ) { +log_stderr ("failure: get_userns_fd" ); + goto out; + } + + open_tree_fd =sys_open_tree (t_dir1_fd,"" , +AT_EMPTY_PATH | +AT_NO_AUTOMOUNT | +AT_SYMLINK_NOFOLLOW | +OPEN_TREE_CLOEXEC | +OPEN_TREE_CLONE ); +if (open_tree_fd <0 ) { +log_stderr ("failure: sys_open_tree" ); + goto out; + } + +if (sys_mount_setattr (open_tree_fd,"" ,AT_EMPTY_PATH , &attr,sizeof (attr))) { +log_stderr ("failure: sys_mount_setattr" ); + goto out; + } + + file1_fd2 =openat (open_tree_fd,FILE1 ,O_RDWR |O_CLOEXEC ,0 ); +if (file1_fd2 <0 ) { +log_stderr ("failure: openat" ); + goto out; + } + +/* +* Verify we can set an v3 fscap for real root this was regressed at +* some point. Make sure this doesn't happen again! +*/ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (!switch_userns (attr.userns_fd,0 ,0 , false)) +die ("failure: switch_userns" ); + +if (expected_dummy_vfs_caps_uid (file1_fd2,-1 )) +die ("failure: expected_dummy_vfs_caps_uid" ); +if (errno !=ENODATA ) +die ("failure: errno" ); + +if (set_dummy_vfs_caps (file1_fd2,0 ,0 )) +die ("failure: set_dummy_vfs_caps" ); + +if (!expected_dummy_vfs_caps_uid (file1_fd2,0 )) +die ("failure: expected_dummy_vfs_caps_uid" ); + +if (!expected_dummy_vfs_caps_uid (file1_fd,0 ) && errno !=EOVERFLOW ) +die ("failure: expected_dummy_vfs_caps_uid" ); + +exit (EXIT_SUCCESS ); + } + +if (wait_for_pid (pid)) + goto out; + +if (!expected_dummy_vfs_caps_uid (file1_fd2,10000 )) { +log_stderr ("failure: expected_dummy_vfs_caps_uid" ); + goto out; + } + +if (!expected_dummy_vfs_caps_uid (file1_fd,0 )) { +log_stderr ("failure: expected_dummy_vfs_caps_uid" ); + goto out; + } + + fret =0 ; +log_debug ("Ran test" ); +out: +safe_close (attr.userns_fd); +safe_close (file1_fd); +safe_close (file1_fd2); +safe_close (open_tree_fd); + +return fret; +} +static int fscaps_idmapped_mounts_in_userns_separate_userns (void ) {int fret =-1 ;
@@ -8717,24 +8832,26 @@ static void usage(void)fprintf (stderr," Run idmapped mount tests\n\n" );fprintf (stderr,"Arguments:\n" ); -fprintf (stderr,"--device Device used in the tests\n" ); -fprintf (stderr,"--fstype Filesystem type used in the tests\n" ); -fprintf (stderr,"--help Print help\n" ); -fprintf (stderr,"--mountpoint Mountpoint of device\n" ); -fprintf (stderr,"--supported Test whether idmapped mounts are supported on this filesystem\n" ); -fprintf (stderr,"--test-core Run core idmapped mount testsuite\n" ); +fprintf (stderr,"--device Device used in the tests\n" ); +fprintf (stderr,"--fstype Filesystem type used in the tests\n" ); +fprintf (stderr,"--help Print help\n" ); +fprintf (stderr,"--mountpoint Mountpoint of device\n" ); +fprintf (stderr,"--supported Test whether idmapped mounts are supported on this filesystem\n" ); +fprintf (stderr,"--test-core Run core idmapped mount testsuite\n" ); +fprintf (stderr,"--test-fscaps-regression Run fscap regression tests\n" );_exit (EXIT_SUCCESS ); }static const struct option longopts[] = { - {"device" , required_argument,0 ,'d' }, - {"fstype" , required_argument,0 ,'f' }, - {"mountpoint" , required_argument,0 ,'m' }, - {"supported" , no_argument,0 ,'s' }, - {"help" , no_argument,0 ,'h' }, - {"test-core" , no_argument,0 ,'c' }, - {NULL ,0 ,0 ,0 }, + {"device" , required_argument,0 ,'d' }, + {"fstype" , required_argument,0 ,'f' }, + {"mountpoint" , required_argument,0 ,'m' }, + {"supported" , no_argument,0 ,'s' }, + {"help" , no_argument,0 ,'h' }, + {"test-core" , no_argument,0 ,'c' }, + {"test-fscaps-regression" , no_argument,0 ,'g' }, + {NULL ,0 ,0 ,0 }, };struct t_idmapped_mounts {
@@ -8748,7 +8865,7 @@ struct t_idmapped_mounts { { fscaps, "fscaps on regular mounts", }, { fscaps_idmapped_mounts, "fscaps on idmapped mounts", }, { fscaps_idmapped_mounts_in_userns, "fscaps on idmapped mounts in user namespace", }, - { fscaps_idmapped_mounts_in_userns_separate_userns, "fscaps on idmapped mounts in user namespace with different id mappings ", }, + { fscaps_idmapped_mounts_in_userns_separate_userns, "fscaps on idmapped mounts in user namespace with different id mappings", }, { fsids_mapped, "mapped fsids", }, { fsids_unmapped, "unmapped fsids", }, { hardlink_crossing_mounts, "cross mount hardlink", },
@@ -8792,6 +8909,10 @@ struct t_idmapped_mounts { { threaded_idmapped_mount_interactions, "threaded operations on idmapped mounts", }, }; +struct t_idmapped_mounts fscaps_in_ancestor_userns[] = { + { fscaps_idmapped_mounts_in_userns_valid_in_ancestor_userns,"fscaps on idmapped mounts in user namespace writing fscap valid in ancestor userns" , }, +}; +static bool run_test (struct t_idmapped_mounts suite[],size_t suite_size) {int i;
@@ -8829,7 +8950,7 @@ int main(int argc, char *argv[]) {int fret, ret;int index =0 ; -bool supported = false, test_core = false; +bool supported = false, test_core = false, test_fscaps_regression = false;while ((ret =getopt_long_only (argc, argv,"" , longopts, &index)) !=-1 ) {switch (ret) {
@@ -8848,6 +8969,9 @@ int main(int argc, char *argv[])case 'c' : test_core = true;break ; +case 'g' : + test_fscaps_regression = true; +break ;case 'h' :/* fallthrough */ default :
@@ -8920,6 +9044,11 @@ int main(int argc, char *argv[])if (test_core && !run_test (basic_suite,ARRAY_SIZE (basic_suite))) goto out; +if (test_fscaps_regression && + !run_test (fscaps_in_ancestor_userns, +ARRAY_SIZE (fscaps_in_ancestor_userns))) + goto out; + fret =EXIT_SUCCESS ; out:
tests/generic/640 new file +28 -0
@@ -0,0 +1,28 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2021 Christian Brauner. All Rights Reserved. +# +# FS QA Test 640 +# +# Test that fscaps on idmapped mounts behave correctly. +# +. ./common/preamble +_begin_fstest auto quick cap idmapped mount + +# get standard environment, filters and checks +. ./common/rc +. ./common/filter + +# real QA test starts here + +_supported_fs generic +_require_idmapped_mounts +_require_test + +echo "Silence is golden" + +$here/src/idmapped-mounts/idmapped-mounts --test-fscaps-regression \ + --device "$TEST_DEV" --mount "$TEST_DIR" --fstype "$FSTYP" + +status=$? +exit
tests/generic/640.out new file +2 -0
@@ -0,0 +1,2 @@ +QA output created by 640 +Silence is golden
Signature · 1 line
2.30.2
PATCH v4 5/8 idmapped-mounts: refactor helpers
In reply to Christian BraunerFrom: Christian Brauner <[email protected]>
Make all userns creation helpers share a commond codebase and move a bunch of code into utils.{c,h}. This simplifies a bunch of things and makes it easier to create nested user namespaces in follow up patches.
---
/* v2 */
unchanged
/* v3 */
unchanged
/* v4 */
unchanged
--- src/idmapped-mounts/mount-idmapped.c | 197 ------------------------- src/idmapped-mounts/utils.c | 209 +++++++++++++++++++-------- src/idmapped-mounts/utils.h | 73 +++++++++- 3 files changed, 223 insertions(+), 256 deletions(-)
src/idmapped-mounts/mount-idmapped.c +0 -197
@@ -27,77 +27,6 @@#include "missing.h" #include "utils.h" -/* A few helpful macros. */ -#define STRLITERALLEN (x) (sizeof(""x"") - 1) - -#define INTTYPE_TO_STRLEN (type) \ - (2 + (sizeof(type) <= 1 \ - ? 3 \ - : sizeof(type) <= 2 \ - ? 5 \ - : sizeof(type) <= 4 \ - ? 10 \ - : sizeof(type) <= 8 ? 20 : sizeof(int[-2 * (sizeof(type) > 8)]))) - -#define syserror (format, ...) \ - ({ \ - fprintf(stderr, format, ##__VA_ARGS__); \ - (-errno); \ - }) - -#define syserror_set (__ret__, format, ...) \ - ({ \ - typeof(__ret__) __internal_ret__ = (__ret__); \ - errno = labs(__ret__); \ - fprintf(stderr, format, ##__VA_ARGS__); \ - __internal_ret__; \ - }) - -struct list { -void *elem; -struct list *next; -struct list *prev; -}; - -#define list_for_each (__iterator, __list) \ - for (__iterator = (__list)->next; __iterator != __list; __iterator = __iterator->next) - -static inline void list_init (struct list *list) -{ - list->elem =NULL ; - list->next = list->prev = list; -} - -static inline int list_empty (const struct list *list) -{ -return list == list->next; -} - -static inline void __list_add (struct list *new,struct list *prev,struct list *next) -{ - next->prev = new; - new->next = next; - new->prev = prev; - prev->next = new; -} - -static inline void list_add_tail (struct list *head,struct list *list) -{ -__list_add (list, head->prev, head); -} - -typedef enum idmap_type_t { -ID_TYPE_UID , -ID_TYPE_GID -}idmap_type_t ; - -struct id_map { -idmap_type_t map_type; -__u32 nsid; -__u32 hostid; -__u32 range; -}; -static struct list active_map; static int add_map_entry(__u32 id_host,
@@ -166,132 +95,6 @@ static int parse_map(char *map)return 0 ; } -static int write_id_mapping (idmap_type_t map_type,pid_t pid,const char *buf,size_t buf_size) -{ -int fd = -EBADF , setgroups_fd = -EBADF ; -int fret =-1 ; -int ret; -char path[STRLITERALLEN ("/proc/" ) +INTTYPE_TO_STRLEN (pid_t) + -STRLITERALLEN ("/setgroups" ) +1 ]; - -if (geteuid () !=0 && map_type ==ID_TYPE_GID ) { - ret =snprintf (path,sizeof (path),"/proc/%d/setgroups" , pid); -if (ret <0 || ret >=sizeof (path)) - goto out; - - setgroups_fd =open (path,O_WRONLY |O_CLOEXEC ); -if (setgroups_fd <0 && errno !=ENOENT ) { -syserror ("Failed to open \"%s\"" , path); - goto out; - } - -if (setgroups_fd >=0 ) { - ret =write_nointr (setgroups_fd,"deny\n" ,STRLITERALLEN ("deny\n" )); -if (ret !=STRLITERALLEN ("deny\n" )) { -syserror ("Failed to write \"deny\" to \"/proc/%d/setgroups\"" , pid); - goto out; - } - } - } - - ret =snprintf (path,sizeof (path),"/proc/%d/%cid_map" , pid, map_type ==ID_TYPE_UID ?'u' :'g' ); -if (ret <0 || ret >=sizeof (path)) - goto out; - - fd =open (path,O_WRONLY |O_CLOEXEC ); -if (fd <0 ) { -syserror ("Failed to open \"%s\"" , path); - goto out; - } - - ret =write_nointr (fd, buf, buf_size); -if (ret != buf_size) { -syserror ("Failed to write %cid mapping to \"%s\"" , - map_type ==ID_TYPE_UID ?'u' :'g' , path); - goto out; - } - - fret =0 ; -out: -if (fd >=0 ) -close (fd); -if (setgroups_fd >=0 ) -close (setgroups_fd); - -return fret; -} - -static int map_ids_from_idmap (struct list *idmap,pid_t pid) -{ -int fill, left; -char mapbuf[4096 ] = {}; -bool had_entry = false; -idmap_type_t map_type, u_or_g; - -for (map_type =ID_TYPE_UID , u_or_g ='u' ; - map_type <=ID_TYPE_GID ; map_type++, u_or_g ='g' ) { -char *pos = mapbuf; -int ret; -struct list *iterator; - - -list_for_each (iterator, idmap) { -struct id_map *map = iterator->elem; -if (map->map_type != map_type) -continue ; - - had_entry = true; - - left =4096 - (pos - mapbuf); - fill =snprintf (pos, left,"%u %u %u\n" , map->nsid, map->hostid, map->range); -/* -* The kernel only takes <= 4k for writes to -* /proc/<pid>/{g,u}id_map -*/ -if (fill <=0 || fill >= left) -return syserror_set (-E2BIG ,"Too many %cid mappings defined" , u_or_g); - - pos += fill; - } -if (!had_entry) -continue ; - - ret =write_id_mapping (map_type, pid, mapbuf, pos - mapbuf); -if (ret <0 ) -return syserror ("Failed to write mapping: %s" , mapbuf); - -memset (mapbuf,0 ,sizeof (mapbuf)); - } - -return 0 ; -} - -static int get_userns_fd_from_idmap (struct list *idmap) -{ -int ret; -pid_t pid; -char path_ns[STRLITERALLEN ("/proc/" ) +INTTYPE_TO_STRLEN (pid_t) + -STRLITERALLEN ("/ns/user" ) +1 ]; - - pid =do_clone (get_userns_fd_cb,NULL ,CLONE_NEWUSER |CLONE_NEWNS ); -if (pid <0 ) -return -errno; - - ret =map_ids_from_idmap (idmap, pid); -if (ret <0 ) -return ret; - - ret =snprintf (path_ns,sizeof (path_ns),"/proc/%d/ns/user" , pid); -if (ret <0 || (size_t )ret >=sizeof (path_ns)) - ret = -EIO ; -else - ret =open (path_ns,O_RDONLY |O_CLOEXEC |O_NOCTTY ); - - (void )kill (pid,SIGKILL ); - (void )wait_for_pid (pid); -return ret; -} -static inline bool strnequal (const char *str,const char *eq,size_t len) {return strncmp (str, eq, len) ==0 ;
src/idmapped-mounts/utils.c +151 -58
250 lines not on this page · raw message
src/idmapped-mounts/utils.h +72 -1
82 lines not on this page · raw message
Signature · 1 line
2.30.2
PATCH v4 6/8 idmapped-mounts: add nested userns creation helpers
In reply to Christian BraunerFrom: Christian Brauner <[email protected]>
Add a helper to create a nested userns hierarchy. This will be used in follow-up tests.
---
/* v2 */
unchanged
/* v3 */
unchanged
/* v4 */
unchanged
--- src/idmapped-mounts/idmapped-mounts.c | 14 --- src/idmapped-mounts/mount-idmapped.c | 32 +----- src/idmapped-mounts/utils.c | 160 +++++++++++++++++++++++++- src/idmapped-mounts/utils.h | 25 ++++ 4 files changed, 185 insertions(+), 46 deletions(-)
src/idmapped-mounts/idmapped-mounts.c +0 -14
@@ -388,20 +388,6 @@ static inline bool switch_fsids(uid_t fsuid, gid_t fsgid)return true; } -static inline bool switch_ids (uid_t uid,gid_t gid) -{ -if (setgroups (0 ,NULL )) -return log_errno (false,"failure: setgroups" ); - -if (setresgid (gid, gid, gid)) -return log_errno (false,"failure: setresgid" ); - -if (setresuid (uid, uid, uid)) -return log_errno (false,"failure: setresuid" ); - -return true; -} -static inline bool switch_userns (int fd,uid_t uid,gid_t gid,bool drop_caps) {if (setns (fd,CLONE_NEWUSER ))
src/idmapped-mounts/mount-idmapped.c +1 -31
@@ -29,36 +29,6 @@static struct list active_map; -static int add_map_entry (__u32 id_host, -__u32 id_ns, -__u32 range, -idmap_type_t map_type) -{ -struct list *new_list =NULL ; -struct id_map *newmap =NULL ; - - newmap =malloc (sizeof (*newmap)); -if (!newmap) -return -ENOMEM ; - - new_list =malloc (sizeof (struct list )); -if (!new_list) { -free (newmap); -return -ENOMEM ; - } - - *newmap = (struct id_map ){ - .hostid = id_host, - .nsid = id_ns, - .range = range, - .map_type = map_type, - }; - - new_list->elem = newmap; -list_add_tail (&active_map, new_list); -return 0 ; -} -static int parse_map (char *map) {char types[2 ] = {'u' ,'g' };
@@ -87,7 +57,7 @@ static int parse_map(char *map)else map_type =ID_TYPE_GID ; - ret =add_map_entry (id_host, id_ns, range, map_type); + ret =add_map_entry (&active_map, id_host, id_ns, range, map_type);if (ret <0 )return ret; }
src/idmapped-mounts/utils.c +159 -1
179 lines not on this page · raw message
src/idmapped-mounts/utils.h +25 -0
@@ -10,6 +10,7 @@#include <linux/types.h> #include <sched.h> #include <signal.h> +#include <stdbool.h> #include <stdio.h> #include <stdlib.h> #include <string.h>
@@ -19,6 +20,9 @@#include "missing.h" +/* Maximum number of nested user namespaces in the kernel. */ +#define MAX_USERNS_LEVEL 32 +/* A few helpful macros. */ #define STRLITERALLEN (x) (sizeof(""x"") - 1)
@@ -63,6 +67,13 @@ struct list {struct list *prev; }; +struct userns_hierarchy { +int fd_userns; +int fd_event; +unsigned int level; +struct list id_map; +}; +#define list_for_each (__iterator, __list) \ for (__iterator = (__list)->next; __iterator != __list; __iterator = __iterator->next)
@@ -90,6 +101,16 @@ static inline void list_add_tail(struct list *head, struct list *list)__list_add (list, head->prev, head); } +static inline void list_del (struct list *list) +{ +struct list *next, *prev; + + next = list->next; + prev = list->prev; + next->prev = prev; + prev->next = next; +} +extern pid_t do_clone (int (*fn)(void *),void *arg,int flags);extern int get_userns_fd (unsigned long nsid,unsigned long hostid,unsigned long range);
@@ -97,5 +118,9 @@ extern int get_userns_fd_from_idmap(struct list *idmap);extern ssize_t read_nointr (int fd,void *buf,size_t count);extern int wait_for_pid (pid_t pid);extern ssize_t write_nointr (int fd,const void *buf,size_t count); +extern bool switch_ids (uid_t uid,gid_t gid); +extern int create_userns_hierarchy (struct userns_hierarchy *h); +extern int add_map_entry (struct list *head,__u32 id_host,__u32 id_ns, +__u32 range,idmap_type_t map_type);#endif /* __IDMAP_UTILS_H */
Signature · 1 line
2.30.2
PATCH v4 7/8 generic/641: add nested user namespace tests
In reply to Christian BraunerFrom: Christian Brauner <[email protected]>
Test ownership and ownership changes in a complex user namespace hierarchy.
---
/* v2 */
- Eryu Guan <[email protected]>:
- Don't create symlinks for each test. Instead, use the same binary and introduce new options to specify which tests to run.
/* v3 */
unchanged
/* v4 */
unchanged
--- src/idmapped-mounts/idmapped-mounts.c | 724 +++++++++++++++++++++++++- src/idmapped-mounts/utils.h | 4 + tests/generic/641 | 28 + tests/generic/641.out | 2 + 4 files changed, 757 insertions(+), 1 deletion(-) create mode 100755 tests/generic/641 create mode 100644 tests/generic/641.out
src/idmapped-mounts/idmapped-mounts.c +723 -1
@@ -8812,6 +8812,714 @@ out:return fret; } +static int nested_userns (void ) +{ +int fret =-1 ; +int ret; +pid_t pid; +struct list *it, *next; +struct userns_hierarchy hierarchy[] = { + { .level =1 , .fd_userns = -EBADF , }, + { .level =2 , .fd_userns = -EBADF , }, + { .level =3 , .fd_userns = -EBADF , }, + { .level =4 , .fd_userns = -EBADF , }, +/* Dummy entry that marks the end. */ + { .level =MAX_USERNS_LEVEL , .fd_userns = -EBADF , }, + }; +struct mount_attr attr_level1 = { + .attr_set =MOUNT_ATTR_IDMAP , + .userns_fd = -EBADF , + }; +struct mount_attr attr_level2 = { + .attr_set =MOUNT_ATTR_IDMAP , + .userns_fd = -EBADF , + }; +struct mount_attr attr_level3 = { + .attr_set =MOUNT_ATTR_IDMAP , + .userns_fd = -EBADF , + }; +struct mount_attr attr_level4 = { + .attr_set =MOUNT_ATTR_IDMAP , + .userns_fd = -EBADF , + }; +int fd_dir1 = -EBADF , + fd_open_tree_level1 = -EBADF , + fd_open_tree_level2 = -EBADF , + fd_open_tree_level3 = -EBADF , + fd_open_tree_level4 = -EBADF ; +const unsigned int id_file_range =10000 ; + +list_init (&hierarchy[0 ].id_map); +list_init (&hierarchy[1 ].id_map); +list_init (&hierarchy[2 ].id_map); +list_init (&hierarchy[3 ].id_map); + +/* +* Give a large map to the outermost user namespace so we can create +* comfortable nested maps. +*/ + ret =add_map_entry (&hierarchy[0 ].id_map,1000000 ,0 ,1000000000 ,ID_TYPE_UID ); +if (ret) { +log_stderr ("failure: adding uidmap for userns at level 1" ); + goto out; + } + + ret =add_map_entry (&hierarchy[0 ].id_map,1000000 ,0 ,1000000000 ,ID_TYPE_GID ); +if (ret) { +log_stderr ("failure: adding gidmap for userns at level 1" ); + goto out; + } + +/* This is uid:0->2000000:100000000 in init userns. */ + ret =add_map_entry (&hierarchy[1 ].id_map,1000000 ,0 ,100000000 ,ID_TYPE_UID ); +if (ret) { +log_stderr ("failure: adding uidmap for userns at level 2" ); + goto out; + } + +/* This is gid:0->2000000:100000000 in init userns. */ + ret =add_map_entry (&hierarchy[1 ].id_map,1000000 ,0 ,100000000 ,ID_TYPE_GID ); +if (ret) { +log_stderr ("failure: adding gidmap for userns at level 2" ); + goto out; + } + +/* This is uid:0->3000000:999 in init userns. */ + ret =add_map_entry (&hierarchy[2 ].id_map,1000000 ,0 ,999 ,ID_TYPE_UID ); +if (ret) { +log_stderr ("failure: adding uidmap for userns at level 3" ); + goto out; + } + +/* This is gid:0->3000000:999 in the init userns. */ + ret =add_map_entry (&hierarchy[2 ].id_map,1000000 ,0 ,999 ,ID_TYPE_GID ); +if (ret) { +log_stderr ("failure: adding gidmap for userns at level 3" ); + goto out; + } + +/* id 999 will remain unmapped. */ + +/* This is uid:1000->2001000:1 in init userns. */ + ret =add_map_entry (&hierarchy[2 ].id_map,1000 ,1000 ,1 ,ID_TYPE_UID ); +if (ret) { +log_stderr ("failure: adding uidmap for userns at level 3" ); + goto out; + } + +/* This is gid:1000->2001000:1 in init userns. */ + ret =add_map_entry (&hierarchy[2 ].id_map,1000 ,1000 ,1 ,ID_TYPE_GID ); +if (ret) { +log_stderr ("failure: adding gidmap for userns at level 3" ); + goto out; + } + +/* This is uid:1001->3001001:10000 in init userns. */ + ret =add_map_entry (&hierarchy[2 ].id_map,1001001 ,1001 ,10000000 ,ID_TYPE_UID ); +if (ret) { +log_stderr ("failure: adding uidmap for userns at level 3" ); + goto out; + } + +/* This is gid:1001->3001001:10000 in init userns. */ + ret =add_map_entry (&hierarchy[2 ].id_map,1001001 ,1001 ,10000000 ,ID_TYPE_GID ); +if (ret) { +log_stderr ("failure: adding gidmap for userns at level 3" ); + goto out; + } + +/* Don't write a mapping in the 4th userns. */ +list_empty (&hierarchy[4 ].id_map); + +/* Create the actual userns hierarchy. */ + ret =create_userns_hierarchy (hierarchy); +if (ret) { +log_stderr ("failure: create userns hierarchy" ); + goto out; + } + + attr_level1.userns_fd = hierarchy[0 ].fd_userns; + attr_level2.userns_fd = hierarchy[1 ].fd_userns; + attr_level3.userns_fd = hierarchy[2 ].fd_userns; + attr_level4.userns_fd = hierarchy[3 ].fd_userns; + +/* +* Create one directory where we create files for each uid/gid within +* the first userns. +*/ +if (mkdirat (t_dir1_fd,DIR1 ,0777 )) { +log_stderr ("failure: mkdirat" ); + goto out; + } + + fd_dir1 =openat (t_dir1_fd,DIR1 ,O_DIRECTORY |O_CLOEXEC ); +if (fd_dir1 <0 ) { +log_stderr ("failure: openat" ); + goto out; + } + +for (unsigned int id =0 ; id <= id_file_range; id++) { +char file[256 ]; + +snprintf (file,sizeof (file),DIR1 "/" FILE1 "_%u" , id); + +if (mknodat (t_dir1_fd, file,S_IFREG |0644 ,0 )) { +log_stderr ("failure: create %s" , file); + goto out; + } + +if (fchownat (t_dir1_fd, file, id, id,AT_SYMLINK_NOFOLLOW )) { +log_stderr ("failure: fchownat %s" , file); + goto out; + } + +if (!expected_uid_gid (t_dir1_fd, file,0 , id, id)) { +log_stderr ("failure: check ownership %s" , file); + goto out; + } + } + +/* Create detached mounts for all the user namespaces. */ + fd_open_tree_level1 =sys_open_tree (t_dir1_fd,DIR1 , +AT_NO_AUTOMOUNT | +AT_SYMLINK_NOFOLLOW | +OPEN_TREE_CLOEXEC | +OPEN_TREE_CLONE ); +if (fd_open_tree_level1 <0 ) { +log_stderr ("failure: sys_open_tree" ); + goto out; + } + + fd_open_tree_level2 =sys_open_tree (t_dir1_fd,DIR1 , +AT_NO_AUTOMOUNT | +AT_SYMLINK_NOFOLLOW | +OPEN_TREE_CLOEXEC | +OPEN_TREE_CLONE ); +if (fd_open_tree_level2 <0 ) { +log_stderr ("failure: sys_open_tree" ); + goto out; + } + + fd_open_tree_level3 =sys_open_tree (t_dir1_fd,DIR1 , +AT_NO_AUTOMOUNT | +AT_SYMLINK_NOFOLLOW | +OPEN_TREE_CLOEXEC | +OPEN_TREE_CLONE ); +if (fd_open_tree_level3 <0 ) { +log_stderr ("failure: sys_open_tree" ); + goto out; + } + + fd_open_tree_level4 =sys_open_tree (t_dir1_fd,DIR1 , +AT_NO_AUTOMOUNT | +AT_SYMLINK_NOFOLLOW | +OPEN_TREE_CLOEXEC | +OPEN_TREE_CLONE ); +if (fd_open_tree_level4 <0 ) { +log_stderr ("failure: sys_open_tree" ); + goto out; + } + +/* Turn detached mounts into detached idmapped mounts. */ +if (sys_mount_setattr (fd_open_tree_level1,"" ,AT_EMPTY_PATH , + &attr_level1,sizeof (attr_level1))) { +log_stderr ("failure: sys_mount_setattr" ); + goto out; + } + +if (sys_mount_setattr (fd_open_tree_level2,"" ,AT_EMPTY_PATH , + &attr_level2,sizeof (attr_level2))) { +log_stderr ("failure: sys_mount_setattr" ); + goto out; + } + +if (sys_mount_setattr (fd_open_tree_level3,"" ,AT_EMPTY_PATH , + &attr_level3,sizeof (attr_level3))) { +log_stderr ("failure: sys_mount_setattr" ); + goto out; + } + +if (sys_mount_setattr (fd_open_tree_level4,"" ,AT_EMPTY_PATH , + &attr_level4,sizeof (attr_level4))) { +log_stderr ("failure: sys_mount_setattr" ); + goto out; + } + +/* Verify that ownership looks correct for callers in the init userns. */ +for (unsigned int id =0 ; id <= id_file_range; id++) { +bool bret; +unsigned int id_level1, id_level2, id_level3; +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + + id_level1 = id +1000000 ; +if (!expected_uid_gid (fd_open_tree_level1, file,0 , id_level1, id_level1)) { +log_stderr ("failure: check ownership %s" , file); + goto out; + } + + id_level2 = id +2000000 ; +if (!expected_uid_gid (fd_open_tree_level2, file,0 , id_level2, id_level2)) { +log_stderr ("failure: check ownership %s" , file); + goto out; + } + +if (id ==999 ) { +/* This id is unmapped. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid); + }else if (id ==1000 ) { + id_level3 = id +2000000 ;/* We punched a hole in the map at 1000. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + }else { + id_level3 = id +3000000 ;/* Rest is business as usual. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + } +if (!bret) { +log_stderr ("failure: check ownership %s" , file); + goto out; + } + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) { +log_stderr ("failure: check ownership %s" , file); + goto out; + } + } + +/* Verify that ownership looks correct for callers in the first userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (!switch_userns (attr_level1.userns_fd,0 ,0 , false)) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +bool bret; +unsigned int id_level1, id_level2, id_level3; +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + + id_level1 = id; +if (!expected_uid_gid (fd_open_tree_level1, file,0 , id_level1, id_level1)) +die ("failure: check ownership %s" , file); + + id_level2 = id +1000000 ; +if (!expected_uid_gid (fd_open_tree_level2, file,0 , id_level2, id_level2)) +die ("failure: check ownership %s" , file); + +if (id ==999 ) { +/* This id is unmapped. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid); + }else if (id ==1000 ) { + id_level3 = id +1000000 ;/* We punched a hole in the map at 1000. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + }else { + id_level3 = id +2000000 ;/* Rest is business as usual. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + } +if (!bret) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + +/* Verify that ownership looks correct for callers in the second userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (!switch_userns (attr_level2.userns_fd,0 ,0 , false)) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +bool bret; +unsigned int id_level2, id_level3; +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + +if (!expected_uid_gid (fd_open_tree_level1, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + + id_level2 = id; +if (!expected_uid_gid (fd_open_tree_level2, file,0 , id_level2, id_level2)) +die ("failure: check ownership %s" , file); + +if (id ==999 ) { +/* This id is unmapped. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid); + }else if (id ==1000 ) { + id_level3 = id;/* We punched a hole in the map at 1000. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + }else { + id_level3 = id +1000000 ;/* Rest is business as usual. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + } +if (!bret) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + +/* Verify that ownership looks correct for callers in the third userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (!switch_userns (attr_level3.userns_fd,0 ,0 , false)) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +bool bret; +unsigned int id_level2, id_level3; +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + +if (!expected_uid_gid (fd_open_tree_level1, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (id ==1000 ) { +/* +* The idmapping of the third userns has a hole +* at uid/gid 1000. That means: +* - 1000->userns_0(2000000) // init userns +* - 1000->userns_1(2000000) // level 1 +* - 1000->userns_2(1000000) // level 2 +* - 1000->userns_3(1000) // level 3 (because level 3 has a hole) +*/ + id_level2 = id; + bret =expected_uid_gid (fd_open_tree_level2, file,0 , id_level2, id_level2); + }else { + bret =expected_uid_gid (fd_open_tree_level2, file,0 , t_overflowuid, t_overflowgid); + } +if (!bret) +die ("failure: check ownership %s" , file); + + +if (id ==999 ) { +/* This id is unmapped. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid); + }else { + id_level3 = id;/* Rest is business as usual. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + } +if (!bret) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + +/* Verify that ownership looks correct for callers in the fourth userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (setns (attr_level4.userns_fd,CLONE_NEWUSER )) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + +if (!expected_uid_gid (fd_open_tree_level1, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level2, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + +/* Verify that chown works correctly for callers in the first userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (!switch_userns (attr_level1.userns_fd,0 ,0 , false)) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +bool bret; +unsigned int id_level1, id_level2, id_level3, id_new; +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + + id_new = id +1 ; +if (fchownat (fd_open_tree_level1, file, id_new, id_new,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + + id_level1 = id_new; +if (!expected_uid_gid (fd_open_tree_level1, file,0 , id_level1, id_level1)) +die ("failure: check ownership %s" , file); + + id_level2 = id_new +1000000 ; +if (!expected_uid_gid (fd_open_tree_level2, file,0 , id_level2, id_level2)) +die ("failure: check ownership %s" , file); + +if (id_new ==999 ) { +/* This id is unmapped. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid); + }else if (id_new ==1000 ) { + id_level3 = id_new +1000000 ;/* We punched a hole in the map at 1000. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + }else { + id_level3 = id_new +2000000 ;/* Rest is business as usual. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + } +if (!bret) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +/* Revert ownership. */ +if (fchownat (fd_open_tree_level1, file, id, id,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + +/* Verify that chown works correctly for callers in the second userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (!switch_userns (attr_level2.userns_fd,0 ,0 , false)) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +bool bret; +unsigned int id_level2, id_level3, id_new; +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + + id_new = id +1 ; +if (fchownat (fd_open_tree_level2, file, id_new, id_new,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + +if (!expected_uid_gid (fd_open_tree_level1, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + + id_level2 = id_new; +if (!expected_uid_gid (fd_open_tree_level2, file,0 , id_level2, id_level2)) +die ("failure: check ownership %s" , file); + +if (id_new ==999 ) { +/* This id is unmapped. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid); + }else if (id_new ==1000 ) { + id_level3 = id_new;/* We punched a hole in the map at 1000. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + }else { + id_level3 = id_new +1000000 ;/* Rest is business as usual. */ + bret =expected_uid_gid (fd_open_tree_level3, file,0 , id_level3, id_level3); + } +if (!bret) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +/* Revert ownership. */ +if (fchownat (fd_open_tree_level2, file, id, id,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + +/* Verify that chown works correctly for callers in the third userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (!switch_userns (attr_level3.userns_fd,0 ,0 , false)) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +unsigned int id_new; +char file[256 ]; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + + id_new = id +1 ; +if (id_new ==999 || id_new ==1000 ) { +/* +* We can't change ownership as we can't +* chown from or to an unmapped id. +*/ +if (!fchownat (fd_open_tree_level3, file, id_new, id_new,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + }else { +if (fchownat (fd_open_tree_level3, file, id_new, id_new,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + } + +if (!expected_uid_gid (fd_open_tree_level1, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +/* There's no id 1000 anymore as we changed ownership for id 1000 to 1001 above. */ +if (!expected_uid_gid (fd_open_tree_level2, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (id_new ==999 ) { +/* +* We did not change ownership as we can't +* chown to an unmapped id. +*/ +if (!expected_uid_gid (fd_open_tree_level3, file,0 , id, id)) +die ("failure: check ownership %s" , file); + }else if (id_new ==1000 ) { +/* +* We did not change ownership as we can't +* chown from an unmapped id. +*/ +if (!expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + }else { +if (!expected_uid_gid (fd_open_tree_level3, file,0 , id_new, id_new)) +die ("failure: check ownership %s" , file); + } + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +/* Revert ownership. */ +if (id_new !=999 && id_new !=1000 ) { +if (fchownat (fd_open_tree_level3, file, id, id,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + } + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + +/* Verify that chown works correctly for callers in the fourth userns. */ + pid =fork (); +if (pid <0 ) { +log_stderr ("failure: fork" ); + goto out; + } +if (pid ==0 ) { +if (setns (attr_level4.userns_fd,CLONE_NEWUSER )) +die ("failure: switch_userns" ); + +for (unsigned int id =0 ; id <= id_file_range; id++) { +char file[256 ]; +unsigned long id_new; + +snprintf (file,sizeof (file),FILE1 "_%u" , id); + + id_new = id +1 ; +if (!fchownat (fd_open_tree_level4, file, id_new, id_new,AT_SYMLINK_NOFOLLOW )) +die ("failure: fchownat %s" , file); + +if (!expected_uid_gid (fd_open_tree_level1, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level2, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level3, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + +if (!expected_uid_gid (fd_open_tree_level4, file,0 , t_overflowuid, t_overflowgid)) +die ("failure: check ownership %s" , file); + + } + +exit (EXIT_SUCCESS ); + } +if (wait_for_pid (pid)) + goto out; + + fret =0 ; +log_debug ("Ran test" ); + +out: +list_for_each_safe (it, &hierarchy[0 ].id_map, next) { +list_del (it); +free (it->elem); +free (it); + } + +list_for_each_safe (it, &hierarchy[1 ].id_map, next) { +list_del (it); +free (it->elem); +free (it); + } + +list_for_each_safe (it, &hierarchy[2 ].id_map, next) { +list_del (it); +free (it->elem); +free (it); + } + +safe_close (hierarchy[0 ].fd_userns); +safe_close (hierarchy[1 ].fd_userns); +safe_close (hierarchy[2 ].fd_userns); +safe_close (fd_dir1); +safe_close (fd_open_tree_level1); +safe_close (fd_open_tree_level2); +safe_close (fd_open_tree_level3); +safe_close (fd_open_tree_level4); +return fret; +} +static void usage (void ) {fprintf (stderr,"Description:\n" );
@@ -8825,6 +9533,7 @@ static void usage(void)fprintf (stderr,"--supported Test whether idmapped mounts are supported on this filesystem\n" );fprintf (stderr,"--test-core Run core idmapped mount testsuite\n" );fprintf (stderr,"--test-fscaps-regression Run fscap regression tests\n" ); +fprintf (stderr,"--test-nested-userns Run nested userns idmapped mount testsuite\n" );_exit (EXIT_SUCCESS ); }
@@ -8837,6 +9546,7 @@ static const struct option longopts[] = { {"help", no_argument, 0, 'h'}, {"test-core", no_argument, 0, 'c'}, {"test-fscaps-regression", no_argument, 0, 'g'}, + {"test-nested-userns", no_argument, 0, 'n'}, {NULL, 0, 0, 0}, };
@@ -8899,6 +9609,10 @@ struct t_idmapped_mounts fscaps_in_ancestor_userns[] = { { fscaps_idmapped_mounts_in_userns_valid_in_ancestor_userns, "fscaps on idmapped mounts in user namespace writing fscap valid in ancestor userns", }, }; +struct t_idmapped_mounts t_nested_userns[] = { + { nested_userns,"test that nested user namespaces behave correctly when attached to idmapped mounts" , }, +}; +static bool run_test (struct t_idmapped_mounts suite[],size_t suite_size) {int i;
@@ -8936,7 +9650,8 @@ int main(int argc, char *argv[]) {int fret, ret;int index =0 ; -bool supported = false, test_core = false, test_fscaps_regression = false; +bool supported = false, test_core = false, + test_fscaps_regression = false, test_nested_userns = false;while ((ret =getopt_long_only (argc, argv,"" , longopts, &index)) !=-1 ) {switch (ret) {
@@ -8958,6 +9673,9 @@ int main(int argc, char *argv[])case 'g' : test_fscaps_regression = true;break ; +case 'n' : + test_nested_userns = true; +break ;case 'h' :/* fallthrough */ default :
@@ -9035,6 +9753,10 @@ int main(int argc, char *argv[]) ARRAY_SIZE(fscaps_in_ancestor_userns))) goto out; +if (test_nested_userns && + !run_test (t_nested_userns,ARRAY_SIZE (t_nested_userns))) + goto out; + fret =EXIT_SUCCESS ; out:
src/idmapped-mounts/utils.h +4 -0
@@ -77,6 +77,10 @@ struct userns_hierarchy {#define list_for_each (__iterator, __list) \ for (__iterator = (__list)->next; __iterator != __list; __iterator = __iterator->next) +#define list_for_each_safe (__iterator, __list, __next) \ + for (__iterator = (__list)->next, __next = __iterator->next; \ + __iterator != __list; __iterator = __next, __next = __next->next) +static inline void list_init (struct list *list) { list->elem =NULL ;
tests/generic/641 new file +28 -0
@@ -0,0 +1,28 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2021 Christian Brauner. All Rights Reserved. +# +# FS QA Test 641 +# +# Test that idmapped mounts behave correctly with complex user namespaces. +# +. ./common/preamble +_begin_fstest auto quick idmapped mount + +# get standard environment, filters and checks +. ./common/rc +. ./common/filter + +# real QA test starts here + +_supported_fs generic +_require_idmapped_mounts +_require_test + +echo "Silence is golden" + +$here/src/idmapped-mounts/idmapped-mounts --test-nested-userns \ + --device "$TEST_DEV" --mount "$TEST_DIR" --fstype "$FSTYP" + +status=$? +exit
tests/generic/641.out new file +2 -0
@@ -0,0 +1,2 @@ +QA output created by 641 +Silence is golden
Signature · 1 line
2.30.2
PATCH v4 8/8 btrfs/244: introduce btrfs specific idmapped mounts tests
In reply to Christian BraunerFrom: Christian Brauner <[email protected]>
While core vfs functionality that btrfs implements is completely covered by the generic test-suite the btrfs specific ioctls are not. This patch expands the test-suite to cover btrfs specific ioctls that required changes to work on idmapped mounts. We deliberately don't use the libbtrfsutil library as we need to know exactly what ioctl's are issued and we need to be in control of all privileges at all times. This test-suite currently tests:
- BTRFS_IOC_{SNAP,SUBVOL}_CREATE_V2
- subvolume creation on idmapped mounts where the fsids do have a mapping in the superblock
- snapshot creation on idmapped mounts where the fsids do have a mapping in the superblock
- subvolume creation on idmapped mounts where the fsids do not have a mapping in the superblock
- snapshot creation on idmapped mounts where the fsids do not have a mapping in the superblock
- subvolume creation on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock
- snapshot creation on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock
- subvolume creation on idmapped mounts where the caller is located in a user namespace and the fsids do not have a mapping in the superblock
- snapshot creation on idmapped mounts where the caller is located in a user namespace and the fsids do not have a mapping in the superblock
- BTRFS_IOC_SNAP_DESTROY_V2
- subvolume deletion on idmapped mounts where the fsids do have a mapping in the superblock
- snapshot deletion on idmapped mounts where the fsids do have a mapping in the superblock
- subvolume deletion on idmapped mounts where the fsids do not have a mapping in the superblock
- snapshot deletion on idmapped mounts where the fsids do not have a mapping in the superblock
- subvolume deletion on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock
- snapshot deletion on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock
- subvolume deletion on idmapped mounts where the caller is located in a user namespace and the fsids do not have a mapping in the superblock
- snapshot deletion on idmapped mounts where the caller is located in a user namespace and the fsids do not have a mapping in the superblock
- unprivileged subvolume deletion on idmapped mounts where the fsids do have a mapping in the superblock and the filesystem is mounted with "user_subvol_rm_allowed"
- unprivileged snapshot deletion on idmapped mounts where the fsids do have a mapping in the superblock and the filesystem is mounted with "user_subvol_rm_allowed"
- subvolume deletion on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock and the filesystem is mounted with "user_subvol_rm_allowed"
- snapshot deletion on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock and the filesystem is mounted with "user_subvol_rm_allowed"
- BTRFS_IOC_SUBVOL_SETFLAGS
- subvolume flags on idmapped mounts where the fsids do have a mapping in the superblock
- snapshot flags on idmapped mounts where the fsids do have a mapping in the superblock
- subvolume flags on idmapped mounts where the fsids do not have a mapping in the superblock
- snapshot flags on idmapped mounts where the fsids do not have a mapping in the superblock
- subvolume flags on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock
- snapshot flags on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock
- subvolume flags on idmapped mounts where the caller is located in a user namespace and the fsids do not have a mapping in the superblock
- snapshot flags on idmapped mounts where the caller is located in a user namespace and the fsids do not have a mapping in the superblock
- BTRFS_IOC_INO_LOOKUP_USER
- subvolume lookup on idmapped mounts where the fsids do have a mapping in the superblock
- subvolume lookup on idmapped mounts where the fsids do not have a mapping in the superblock
- subvolume lookup on idmapped mounts where the caller is located in a user namespace and the fsids do have a mapping in the superblock
- subvolume lookup on idmapped mounts where the caller is located in a user namespace and the fsids do not have a mapping in the superblock
---
/* v2 */
new patch
/* v3 */
- Christian Brauner <[email protected]>:
- Ensure tests can be compiled on systems that lack the relevant btrfs types and constants.
/* v4 */
unchanged
--- configure.ac | 10 +- src/idmapped-mounts/idmapped-mounts.c | 3846 ++++++++++++++++++++++++- tests/btrfs/244 | 34 + tests/btrfs/244.out | 2 + 4 files changed, 3890 insertions(+), 2 deletions(-) create mode 100755 tests/btrfs/244 create mode 100644 tests/btrfs/244.out
configure.ac +9 -1
@@ -67,9 +67,17 @@ AC_PACKAGE_WANT_LINUX_FS_H AC_PACKAGE_WANT_LIBBTRFSUTIL AC_HAVE_COPY_FILE_RANGE - AC_CHECK_FUNCS([renameat2]) AC_CHECK_TYPES([struct mount_attr], [], [], [[#include <linux/mount.h>]]) +AC_CHECK_TYPES([struct btrfs_qgroup_limit], [], [], [[#include <linux/btrfs.h>]]) +AC_CHECK_TYPES([struct btrfs_qgroup_inherit], [], [], [[#include <linux/btrfs.h>]]) +AC_CHECK_TYPES([struct btrfs_ioctl_vol_args], [], [], [[#include <linux/btrfs.h>]]) +AC_CHECK_TYPES([struct btrfs_ioctl_vol_args_v2], [], [], [[#include <linux/btrfs.h>]]) +AC_CHECK_TYPES([struct btrfs_ioctl_ino_lookup_args], [], [], [[#include <linux/btrfs.h>]]) +AC_CHECK_TYPES([struct btrfs_ioctl_ino_lookup_user_args], [], [], [[#include <linux/btrfs.h>]]) +AC_CHECK_TYPES([struct btrfs_ioctl_get_subvol_rootref_args], [], [], [[#include <linux/btrfs.h>]]) +AC_CHECK_HEADERS([linux/btrfs.h linux/btrfs_tree.h]) + AC_CONFIG_HEADER(include/config.h) AC_CONFIG_FILES([include/builddefs])
src/idmapped-mounts/idmapped-mounts.c +3845 -1
2,982 lines not on this page · raw message
The rest of this diff is not shown. View the raw message.
Signature · 1 line
2.30.2
Part of the diff is not shown here. View the raw message.
Re: [PATCH v4 2/8] idmapped-mounts: switch to getopt_long_only()
In reply to Christian BraunerLooks good,
Quoted 29 lines · Christian Brauner
From: Christian Brauner <[email protected]>
Hey everyone,
This is v4 with only a minimal change to
getopt_long_only(). The rest is identical. In v3 the changes requested by Christoph and his Reviewed-bys added. I've also added Josef's Reviewed-by for the newly added btrfs specific idmapped mount testsuite. There are no major changes to v2.This time around I've put everyone in To: to make sure that they receive all patches. The list apparently still refues patches if they are fairly huge. The series can be pulled from three locations:
[email protected]:pub/scm/linux/kernel/git/brauner/xfstests-dev fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns
This adds three new tests:
- a regression test for vfs capabilities
- a new test with nested and complex idmapping layouts
- a new btrfs specific idmapped mount testsuiteSince v2 patches to support idmapped mounts on btrfs have been merged into btrfs' for-next tree:
https://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux.git/commit/?h=for-next&id=f691ae1e0bc591df5e9897387d726adbb440fc4c
and so are on track to be included for v5.15. I'd like to get the btrfs tests merged now so linux-next can be tested with the them. I hope that's ok.
Quoted 3 lines
I ran:
git rebase -i -x "make -j4 clean && make -j4 && sudo ./check -g idmapped" HEAD~8 on the whole series. Everthing passes.
Hey Eryu,
Sorry for asking but it's been a few months. Did you miss this series intentionally this iteration? After last weeks mail I added all Acks/Reviewed-bys that came in so this should hopefully be good to go.
I'd really like to see these new tests going in rather soon as their quite crucial to ensure correct functionality especially since we're seeing new filesystems making use of idmapped mounts.
Thanks!
Christian
Quoted 34 lines · Christian Brauner
Quoted 29 lines · Christian Brauner
From: Christian Brauner <[email protected]>
Hey everyone,
This is v4 with only a minimal change to
getopt_long_only(). The rest is identical. In v3 the changes requested by Christoph and his Reviewed-bys added. I've also added Josef's Reviewed-by for the newly added btrfs specific idmapped mount testsuite. There are no major changes to v2.This time around I've put everyone in To: to make sure that they receive all patches. The list apparently still refues patches if they are fairly huge. The series can be pulled from three locations:
[email protected]:pub/scm/linux/kernel/git/brauner/xfstests-dev fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns
This adds three new tests:
- a regression test for vfs capabilities
- a new test with nested and complex idmapping layouts
- a new btrfs specific idmapped mount testsuiteSince v2 patches to support idmapped mounts on btrfs have been merged into btrfs' for-next tree:
https://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux.git/commit/?h=for-next&id=f691ae1e0bc591df5e9897387d726adbb440fc4c
and so are on track to be included for v5.15. I'd like to get the btrfs tests merged now so linux-next can be tested with the them. I hope that's ok.Quoted 3 lines
I ran:
git rebase -i -x "make -j4 clean && make -j4 && sudo ./check -g idmapped" HEAD~8 on the whole series. Everthing passes.
Quoted 5 lines
Hey Eryu,
Sorry for asking but it's been a few months. Did you miss this series intentionally this iteration? After last weeks mail I added all Acks/Reviewed-bys that came in so this should hopefully be good to go.
Thanks for the revision, that really helps! And I left this patchset intentionally, as it's a big patchset that may need more exposure time.
Quoted 4 lines
I'd really like to see these new tests going in rather soon as their quite crucial to ensure correct functionality especially since we're seeing new filesystems making use of idmapped mounts.
Sure, I'll merge them this week if there's no more comments.
Thanks,
Eryu
Quoted 45 lines · Eryu Guan
Quoted 34 lines · Christian Brauner
Quoted 29 lines · Christian Brauner
From: Christian Brauner <[email protected]>
Hey everyone,
This is v4 with only a minimal change to
getopt_long_only(). The rest is identical. In v3 the changes requested by Christoph and his Reviewed-bys added. I've also added Josef's Reviewed-by for the newly added btrfs specific idmapped mount testsuite. There are no major changes to v2.This time around I've put everyone in To: to make sure that they receive all patches. The list apparently still refues patches if they are fairly huge. The series can be pulled from three locations:
[email protected]:pub/scm/linux/kernel/git/brauner/xfstests-dev fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns [email protected]:brauner/xfstests.git fs.idmapped.nested_userns
This adds three new tests:
- a regression test for vfs capabilities
- a new test with nested and complex idmapping layouts
- a new btrfs specific idmapped mount testsuiteSince v2 patches to support idmapped mounts on btrfs have been merged into btrfs' for-next tree:
https://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux.git/commit/?h=for-next&id=f691ae1e0bc591df5e9897387d726adbb440fc4c
and so are on track to be included for v5.15. I'd like to get the btrfs tests merged now so linux-next can be tested with the them. I hope that's ok.Quoted 3 lines
I ran:
git rebase -i -x "make -j4 clean && make -j4 && sudo ./check -g idmapped" HEAD~8 on the whole series. Everthing passes.Quoted 5 lines
Hey Eryu,
Sorry for asking but it's been a few months. Did you miss this series intentionally this iteration? After last weeks mail I added all Acks/Reviewed-bys that came in so this should hopefully be good to go.
Thanks for the revision, that really helps! And I left this patchset intentionally, as it's a big patchset that may need more exposure time.
Quoted 1 line
Quoted 5 lines
Quoted 3 lines
I'd really like to see these new tests going in rather soon as their quite crucial to ensure correct functionality especially since we're seeing new filesystems making use of idmapped mounts.
Sure, I'll merge them this week if there's no more comments.
Perfect, thank you!
Christian