openat2: minor uapi cleanups
Subsystems (4)
cover openat2: minor uapi cleanups
While openat2(2) is still not yet in Linus's tree, we can take this opportunity to iron out some small warts that weren't noticed earlier:
* A fix was suggested by Florian Weimer, to separate the openat2 definitions so glibc can use the header directly. I've put the maintainership under VFS but let me know if you'd prefer it belong ot the fcntl folks.
* Having heterogenous field sizes in an extensible struct results in "padding hole" problems when adding new fields (in addition the correct error to use for non-zero padding isn't entirely clear ). The simplest solution is to just copy clone(3)'s model -- always use u64s. It will waste a little more space in the struct, but it removes a possible future headache.
Aleksa Sarai (2):
uapi: split openat2(2) definitions from fcntl.h
openat2: drop open_how->__padding field
MAINTAINERS | 1 + fs/open.c | 2 - include/uapi/linux/fcntl.h | 37 +---------------- include/uapi/linux/openat2.h | 40 +++++++++++++++++++ tools/testing/selftests/openat2/helpers.h | 3 +- .../testing/selftests/openat2/openat2_test.c | 24 ++++------- 6 files changed, 51 insertions(+), 56 deletions(-) create mode 100644 include/uapi/linux/openat2.h
base-commit: 912dfe068c43fa13c587b8d30e73d335c5ba7d44
Signature · 1 line
2.24.0
PATCH 1/2 uapi: split openat2(2) definitions from fcntl.h
Florian mentioned that glibc doesn't use fcntl.h because it has some issues with namespace cleanliness, and that we should have a separate header for openat2(2) if possible.
--- MAINTAINERS | 1 + include/uapi/linux/fcntl.h | 37 +------------------------------- include/uapi/linux/openat2.h | 41 ++++++++++++++++++++++++++++++++++++ 3 files changed, 43 insertions(+), 36 deletions(-) create mode 100644 include/uapi/linux/openat2.h
MAINTAINERS +1 -0
@@ -6397,6 +6397,7 @@ F: fs/* F: include/linux/fs.h F: include/linux/fs_types.h F: include/uapi/linux/fs.h +F: include/uapi/linux/openat2.h FINTEK F75375S HARDWARE MONITOR AND FAN CONTROLLER DRIVER M: Riku Voipio <[email protected]>
include/uapi/linux/fcntl.h +1 -36
@@ -3,6 +3,7 @@#define _UAPI_LINUX_FCNTL_H#include <asm/fcntl.h> +#include <linux/openat2.h> #define F_SETLEASE (F_LINUX_SPECIFIC_BASE + 0)#define F_GETLEASE (F_LINUX_SPECIFIC_BASE + 1)
@@ -100,40 +101,4 @@#define AT_RECURSIVE 0x8000/* Apply to the entire subtree */ -/* -* Arguments for how openat2(2) should open the target path. If @resolve is -* zero, then openat2(2) operates very similarly to openat(2). -* -* However, unlike openat(2), unknown bits in @flags result in -EINVAL rather -* than being silently ignored. @mode must be zero unless one of {O_CREAT, -* O_TMPFILE} are set. -* -* @flags: O_* flags. -* @mode: O_CREAT/O_TMPFILE file mode. -* @resolve: RESOLVE_* flags. -*/ -struct open_how { -__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ -__aligned_u64 resolve; -}; - -#define OPEN_HOW_SIZE_VER0 24/* sizeof first published struct */ -#define OPEN_HOW_SIZE_LATEST OPEN_HOW_SIZE_VER0 - -/* how->resolve flags for openat2(2). */ -#define RESOLVE_NO_XDEV 0x01/* Block mount-point crossings -(includes bind-mounts). */ -#define RESOLVE_NO_MAGICLINKS 0x02/* Block traversal through procfs-style -"magic-links". */ -#define RESOLVE_NO_SYMLINKS 0x04/* Block traversal through all symlinks -(implies OEXT_NO_MAGICLINKS) */ -#define RESOLVE_BENEATH 0x08/* Block "lexical" trickery like -"..", symlinks, and absolute -paths which escape the dirfd. */ -#define RESOLVE_IN_ROOT 0x10/* Make all jumps to "/" and ".." -be scoped inside the dirfd -(similar to chroot(2)). */ -#endif /* _UAPI_LINUX_FCNTL_H */
include/uapi/linux/openat2.h new file +41 -0
41 lines not on this page · raw message
Signature · 1 line
2.24.0
Re: [PATCH 1/2] uapi: split openat2(2) definitions from fcntl.h
* Aleksa Sarai:
include/uapi/linux/openat2.h line 3
Quoted 9 lines
include/uapi/linux/openat2.h new file +3
-0@@ -0,0 +1,41 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _UAPI_LINUX_OPENAT2_H +#define _UAPI_LINUX_OPENAT2_H
I think you should include the relevant header for __align_u64 etc. here.
[…]
Quoted 10 lines
include/uapi/linux/openat2.h new file +10
-0+* Arguments for how openat2(2) should open the target path. If @resolve is +* zero, then openat2(2) operates very similarly to openat(2). +* +* However, unlike openat(2), unknown bits in @flags result in -EINVAL rather +* than being silently ignored. @mode must be zero unless one of {O_CREAT, +* O_TMPFILE} are set. +* +* @flags: O_* flags. +* @mode: O_CREAT/O_TMPFILE file mode. +* @resolve: RESOLVE_* flags.
include/uapi/linux/openat2.h line 25
Quoted 10 lines
include/uapi/linux/openat2.h new file +10
-0+*/ +struct open_how { +__aligned_u64 flags; +__u16 mode; +__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 resolve; +}; + +#define OPEN_HOW_SIZE_VER0 24/* sizeof first published struct */ +#define OPEN_HOW_SIZE_LATEST OPEN_HOW_SIZE_VER0
Are these really useful for the UAPI header? Is there a situation where OPEN_HOW_SIZE_LATEST would be different from sizeof (struct open_how)?
The header is not compatible with the assembler anyway, so the numeric constant does not seem useful.
Thanks,
Florian
Re: [PATCH 1/2] uapi: split openat2(2) definitions from fcntl.h
Quoted 14 lines · Florian Weimer
* Aleksa Sarai:
include/uapi/linux/openat2.h line 3
Quoted 9 lines
include/uapi/linux/openat2.h new file +3
-0@@ -0,0 +1,41 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _UAPI_LINUX_OPENAT2_H +#define _UAPI_LINUX_OPENAT2_HI think you should include the relevant header for
__align_u64etc. here.
Right -- no idea why I forgot to include them.
Quoted 25 lines
[…]
Quoted 11 lines
+ * Arguments for how openat2(2) should open the target path. If @resolve is
+ * zero, then openat2(2) operates very similarly to openat(2).
+ *
+ * However, unlike openat(2), unknown bits in @flags result in-EINVALrather
+ * than being silently ignored. @mode must be zero unless one of {O_CREAT,
+ *O_TMPFILE} are set.
+ *
+ * @flags: O_* flags.
+ * @mode: O_CREAT/O_TMPFILE file mode.
+ * @resolve: RESOLVE_* flags.
+ */include/uapi/linux/openat2.h line 25
Quoted 9 lines
include/uapi/linux/openat2.h new file +9
-0+struct open_how { +__aligned_u64 flags; +__u16 mode; +__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 resolve; +}; + +#define OPEN_HOW_SIZE_VER0 24/* sizeof first published struct */ +#define OPEN_HOW_SIZE_LATEST OPEN_HOW_SIZE_VER0Are these really useful for the UAPI header? Is there a situation where
OPEN_HOW_SIZE_LATESTwould be different from sizeof (structopen_how)?
Quoted 2 lines
The header is not compatible with the assembler anyway, so the numeric constant does not seem useful.
OPEN_HOW_SIZE_VER0 could conceivably be useful (in the future we may do size-based checks) but maybe we can just expose it if someone actually ends up needing it.
I will move them to the in-kernel header (we use them for BUILD_BUG_ONs to make sure that the sizes are correct).
Signature · 4 lines
Aleksa Sarai Senior Software Engineer (Containers) SUSE Linux GmbH <https://www.cyphar.com/>
RE: [PATCH 1/2] uapi: split openat2(2) definitions from fcntl.h
From: Aleksa Sarai
Quoted 17 lines
Sent: 19 December 2019 13:45
Quoted 14 lines · Florian Weimer
* Aleksa Sarai:
include/uapi/linux/openat2.h line 3
Quoted 9 lines
include/uapi/linux/openat2.h new file +3
-0@@ -0,0 +1,41 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _UAPI_LINUX_OPENAT2_H +#define _UAPI_LINUX_OPENAT2_HI think you should include the relevant header for
__align_u64etc. here.
Quoted 1 line
Right -- no idea why I forgot to include them.
I'm guessing that is just 64bit aligned on 32bit archs like x86?
No need to enforce it provided the structure will have no padding on archs where the 64bit fields are 64bit aligned.
A plain __u64 should be fine.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
Re: [PATCH 1/2] uapi: split openat2(2) definitions from fcntl.h
Quoted 18 lines · David Laight
From: Aleksa Sarai
Quoted 17 lines
Sent: 19 December 2019 13:45
Quoted 14 lines · Florian Weimer
* Aleksa Sarai:
include/uapi/linux/openat2.h line 3
Quoted 9 lines
include/uapi/linux/openat2.h new file +3
-0@@ -0,0 +1,41 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _UAPI_LINUX_OPENAT2_H +#define _UAPI_LINUX_OPENAT2_HI think you should include the relevant header for
__align_u64etc. here.
Quoted 3 lines
Quoted 1 line
Right -- no idea why I forgot to include them.
I'm guessing that is just 64bit aligned on 32bit archs like x86?
Yeah,
#define __aligned_u64 __u64 __attribute__((aligned(8)))
Quoted 3 lines
No need to enforce it provided the structure will have no padding on archs where the 64bit fields are 64bit aligned. A plain
__u64should be fine.
Will this cause problems for x86-on-x86_64 emulation? Requiring an 8-byte alignment for 'struct open_how' really isn't that undue of a burden IMHO. Then again, clone3 is a bit of an outlier since both perf_event_open and sched_setattr just use __u64s.
Signature · 4 lines
Aleksa Sarai Senior Software Engineer (Containers) SUSE Linux GmbH <https://www.cyphar.com/>
RE: [PATCH 1/2] uapi: split openat2(2) definitions from fcntl.h
From: Aleksa Sarai
Quoted 1 line
Sent: 20 December 2019 09:32
...
Quoted 6 lines
Quoted 1 line
I'm guessing that is just 64bit aligned on 32bit archs like x86?
Yeah,
#define
__aligned_u64__u64__attribute__((aligned(8)))
Quoted 8 lines
Quoted 3 lines
No need to enforce it provided the structure will have no padding on archs where the 64bit fields are 64bit aligned. A plain
__u64should be fine.Will this cause problems for x86-on-x86_64 emulation? Requiring an 8-byte alignment for 'struct
open_how' really isn't that undue of a burden IMHO. Then again, clone3 is a bit of an outlier since bothperf_event_openandsched_setattrjust use__u64s.
Makes diddly-squit difference.
The 64bit kernel will 64bit align the structure.
The kernel must allow for the userspace structure having arbitrary alignment. So there is no reason to (try to) align the user structure.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
PATCH 1/2 uapi: split openat2(2) definitions from fcntl.h
In reply to Aleksa SaraiFlorian mentioned that glibc doesn't use fcntl.h because it has some issues with namespace cleanliness, and that we should have a separate header for openat2(2) if possible.
--- MAINTAINERS | 1 + include/uapi/linux/fcntl.h | 37 +------------------------------- include/uapi/linux/openat2.h | 41 ++++++++++++++++++++++++++++++++++++ 3 files changed, 43 insertions(+), 36 deletions(-) create mode 100644 include/uapi/linux/openat2.h
MAINTAINERS +1 -0
@@ -6397,6 +6397,7 @@ F: fs/* F: include/linux/fs.h F: include/linux/fs_types.h F: include/uapi/linux/fs.h +F: include/uapi/linux/openat2.h FINTEK F75375S HARDWARE MONITOR AND FAN CONTROLLER DRIVER M: Riku Voipio <[email protected]>
include/uapi/linux/fcntl.h +1 -36
@@ -3,6 +3,7 @@#define _UAPI_LINUX_FCNTL_H#include <asm/fcntl.h> +#include <linux/openat2.h> #define F_SETLEASE (F_LINUX_SPECIFIC_BASE + 0)#define F_GETLEASE (F_LINUX_SPECIFIC_BASE + 1)
@@ -100,40 +101,4 @@#define AT_RECURSIVE 0x8000/* Apply to the entire subtree */ -/* -* Arguments for how openat2(2) should open the target path. If @resolve is -* zero, then openat2(2) operates very similarly to openat(2). -* -* However, unlike openat(2), unknown bits in @flags result in -EINVAL rather -* than being silently ignored. @mode must be zero unless one of {O_CREAT, -* O_TMPFILE} are set. -* -* @flags: O_* flags. -* @mode: O_CREAT/O_TMPFILE file mode. -* @resolve: RESOLVE_* flags. -*/ -struct open_how { -__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ -__aligned_u64 resolve; -}; - -#define OPEN_HOW_SIZE_VER0 24/* sizeof first published struct */ -#define OPEN_HOW_SIZE_LATEST OPEN_HOW_SIZE_VER0 - -/* how->resolve flags for openat2(2). */ -#define RESOLVE_NO_XDEV 0x01/* Block mount-point crossings -(includes bind-mounts). */ -#define RESOLVE_NO_MAGICLINKS 0x02/* Block traversal through procfs-style -"magic-links". */ -#define RESOLVE_NO_SYMLINKS 0x04/* Block traversal through all symlinks -(implies OEXT_NO_MAGICLINKS) */ -#define RESOLVE_BENEATH 0x08/* Block "lexical" trickery like -"..", symlinks, and absolute -paths which escape the dirfd. */ -#define RESOLVE_IN_ROOT 0x10/* Make all jumps to "/" and ".." -be scoped inside the dirfd -(similar to chroot(2)). */ -#endif /* _UAPI_LINUX_FCNTL_H */
include/uapi/linux/openat2.h new file +41 -0
41 lines not on this page · raw message
Signature · 1 line
2.24.0
cover openat2: minor uapi cleanups
In reply to Aleksa SaraiWhile openat2(2) is still not yet in Linus's tree, we can take this opportunity to iron out some small warts that weren't noticed earlier:
* A fix was suggested by Florian Weimer, to separate the openat2 definitions so glibc can use the header directly. I've put the maintainership under VFS but let me know if you'd prefer it belong ot the fcntl folks.
* Having heterogenous field sizes in an extensible struct results in "padding hole" problems when adding new fields (in addition the correct error to use for non-zero padding isn't entirely clear ). The simplest solution is to just copy clone(3)'s model -- always use u64s. It will waste a little more space in the struct, but it removes a possible future headache.
Aleksa Sarai (2):
uapi: split openat2(2) definitions from fcntl.h
openat2: drop open_how->__padding field
MAINTAINERS | 1 + fs/open.c | 2 - include/uapi/linux/fcntl.h | 37 +---------------- include/uapi/linux/openat2.h | 40 +++++++++++++++++++ tools/testing/selftests/openat2/helpers.h | 3 +- .../testing/selftests/openat2/openat2_test.c | 24 ++++------- 6 files changed, 51 insertions(+), 56 deletions(-) create mode 100644 include/uapi/linux/openat2.h
base-commit: 912dfe068c43fa13c587b8d30e73d335c5ba7d44
Signature · 1 line
2.24.0
PATCH 2/2 openat2: drop open_how->__padding field
In reply to Aleksa SaraiThe purpose of explicit padding was to allow us to use the space in the future (C provides no guarantee about the value of padding bytes and thus userspace could've provided garbage).
However, the downside of explicit padding is that any extension we wish to add should fit the space exactly (otherwise we may end up with a u16 which will never be used). In addition, the correct error to return for non-zero padding is not clear (-EINVAL doesn't imply "you're using an extension field unsupported by this kernel", but -E2BIG seems a bit odd if the structure size isn't different).
The simplest solution is to just match the design of clone3(2) -- use u64s for all fields. The extra few-bytes cost of extra fields is not significant (it's unlikely configuration structs will ever be extremely large) and it allows for more flag space if necessary.
As openat2(2) is not yet in Linus's tree, we can iron out these minor warts before we commit to this as a stable ABI.
--- fs/open.c | 2 -- include/uapi/linux/openat2.h | 3 +-- tools/testing/selftests/openat2/helpers.h | 3 +-- .../testing/selftests/openat2/openat2_test.c | 24 +++++++------------ 4 files changed, 10 insertions(+), 22 deletions(-)
fs/open.c +0 -2
@@ -993,8 +993,6 @@ static inline int build_open_flags(const struct open_how *how,return -EINVAL ;if (how->resolve & ~VALID_RESOLVE_FLAGS )return -EINVAL ; -if (memchr_inv (how->__padding,0 ,sizeof (how->__padding))) -return -EINVAL ;/* Deal with the mode. */ if (WILL_CREATE (flags)) {
include/uapi/linux/openat2.h +1 -2
@@ -16,8 +16,7 @@*/ struct open_how {__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 mode;__aligned_u64 resolve; };
tools/testing/selftests/openat2/helpers.h +1 -2
@@ -37,8 +37,7 @@*/ struct open_how {__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 mode;__aligned_u64 resolve; };
tools/testing/selftests/openat2/openat2_test.c +8 -16
@@ -40,7 +40,7 @@ struct struct_test {int err; }; -#define NUM_OPENAT2_STRUCT_TESTS 10 +#define NUM_OPENAT2_STRUCT_TESTS 7#define NUM_OPENAT2_STRUCT_VARIATIONS 13void test_openat2_struct (void )
@@ -57,20 +57,6 @@ void test_openat2_struct(void) .arg.inner.flags = O_RDONLY, .size = sizeof(struct open_how_ext) }, - /* Normal struct with broken padding. */ - { .name = "normal struct (non-zero padding[0])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0xa0, 0x00, 0x00}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, - { .name = "normal struct (non-zero padding[1])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0x00, 0x1a, 0x00}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, - { .name = "normal struct (non-zero padding[2])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0x00, 0x00, 0xef}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, -/* TODO: Once expanded, check zero-padding. */ /* Smaller than version-0 struct. */
@@ -169,7 +155,7 @@ struct flag_test {int err; }; -#define NUM_OPENAT2_FLAG_TESTS 21 +#define NUM_OPENAT2_FLAG_TESTS 23void test_openat2_flags (void ) {
@@ -214,9 +200,15 @@ void test_openat2_flags(void) { .name = "invalid how.mode and O_CREAT", .how.flags = O_CREAT, .how.mode = 0xFFFF, .err = -EINVAL }, + { .name = "invalid (very large) how.mode and O_CREAT", + .how.flags = O_CREAT, + .how.mode = 0xC000000000000000ULL, .err = -EINVAL }, { .name = "invalid how.mode and O_TMPFILE", .how.flags = O_TMPFILE | O_RDWR, .how.mode = 0x1337, .err = -EINVAL }, + { .name = "invalid (very large) how.mode and O_TMPFILE", + .how.flags = O_TMPFILE | O_RDWR, + .how.mode = 0x0000A00000000000ULL, .err = -EINVAL }, /* ->resolve must only contain RESOLVE_* flags. */ { .name = "invalid how.resolve and O_RDONLY",
Signature · 1 line
2.24.0
Re: [PATCH 2/2] openat2: drop open_how->__padding field
Quoted 16 lines · Aleksa Sarai
The purpose of explicit padding was to allow us to use the space in the future (C provides no guarantee about the value of padding bytes and thus userspace could've provided garbage).
However, the downside of explicit padding is that any extension we wish to add should fit the space exactly (otherwise we may end up with a u16 which will never be used). In addition, the correct error to return for non-zero padding is not clear (
-EINVALdoesn't imply "you're using an extension field unsupported by this kernel", but-E2BIGseems a bit odd if the structure size isn't different).The simplest solution is to just match the design of clone3(2) -- use u64s for all fields. The extra few-bytes cost of extra fields is not significant (it's unlikely configuration structs will ever be extremely large) and it allows for more flag space if necessary.
Quoted 5 lines
As openat2(2) is not yet in Linus's tree, we can iron out these minor warts before we commit to this as a stable ABI.
Suggested-by David Laight <[email protected]>Signed-off-by Aleksa Sarai <[email protected]>
Fine with me.
Quoted 110 lines
--- fs/open.c | 2 -- include/uapi/linux/openat2.h | 3 +-- tools/testing/selftests/openat2/helpers.h | 3 +-- .../testing/selftests/openat2/openat2_test.c | 24 +++++++------------ 4 files changed, 10 insertions(+), 22 deletions(-)fs/open.c +0
-2@@ -993,8 +993,6 @@ static inline int build_open_flags(const struct open_how *how,return -EINVAL ;if (how->resolve & ~VALID_RESOLVE_FLAGS )return -EINVAL ; -if (memchr_inv (how->__padding,0 ,sizeof (how->__padding))) -return -EINVAL ;/* Deal with the mode. */ if (WILL_CREATE (flags)) {include/uapi/linux/openat2.h +1
-2@@ -16,8 +16,7 @@*/ struct open_how {__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 mode;__aligned_u64 resolve; };tools/testing/selftests/openat2/helpers.h +1
-2@@ -37,8 +37,7 @@*/ struct open_how {__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 mode;__aligned_u64 resolve; };tools/testing/selftests/openat2/openat2_test.c +8
-16@@ -40,7 +40,7 @@ struct struct_test {int err; }; -#define NUM_OPENAT2_STRUCT_TESTS 10 +#define NUM_OPENAT2_STRUCT_TESTS 7#define NUM_OPENAT2_STRUCT_VARIATIONS 13void test_openat2_struct (void )@@ -57,20 +57,6 @@ void test_openat2_struct(void) .arg.inner.flags = O_RDONLY, .size = sizeof(struct open_how_ext) }, - /* Normal struct with broken padding. */ - { .name = "normal struct (non-zero padding[0])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0xa0, 0x00, 0x00}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, - { .name = "normal struct (non-zero padding[1])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0x00, 0x1a, 0x00}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, - { .name = "normal struct (non-zero padding[2])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0x00, 0x00, 0xef}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, -/* TODO: Once expanded, check zero-padding. */ /* Smaller than version-0 struct. */ @@ -169,7 +155,7 @@ struct flag_test {int err; }; -#define NUM_OPENAT2_FLAG_TESTS 21 +#define NUM_OPENAT2_FLAG_TESTS 23void test_openat2_flags (void ) {@@ -214,9 +200,15 @@ void test_openat2_flags(void) { .name = "invalid how.mode and O_CREAT", .how.flags = O_CREAT, .how.mode = 0xFFFF, .err = -EINVAL }, + { .name = "invalid (very large) how.mode and O_CREAT", + .how.flags = O_CREAT, + .how.mode = 0xC000000000000000ULL, .err = -EINVAL }, { .name = "invalid how.mode and O_TMPFILE", .how.flags = O_TMPFILE | O_RDWR, .how.mode = 0x1337, .err = -EINVAL }, + { .name = "invalid (very large) how.mode and O_TMPFILE", + .how.flags = O_TMPFILE | O_RDWR, + .how.mode = 0x0000A00000000000ULL, .err = -EINVAL }, /* ->resolve must only contain RESOLVE_* flags. */ { .name = "invalid how.resolve and O_RDONLY",Signature · 1 line
2.24.0
PATCH 2/2 openat2: drop open_how->__padding field
In reply to Aleksa SaraiThe purpose of explicit padding was to allow us to use the space in the future (C provides no guarantee about the value of padding bytes and thus userspace could've provided garbage).
However, the downside of explicit padding is that any extension we wish to add should fit the space exactly (otherwise we may end up with a u16 which will never be used). In addition, the correct error to return for non-zero padding is not clear (-EINVAL doesn't imply "you're using an extension field unsupported by this kernel", but -E2BIG seems a bit odd if the structure size isn't different).
The simplest solution is to just match the design of clone3(2) -- use u64s for all fields. The extra few-bytes cost of extra fields is not significant (it's unlikely configuration structs will ever be extremely large) and it allows for more flag space if necessary.
As openat2(2) is not yet in Linus's tree, we can iron out these minor warts before we commit to this as a stable ABI.
--- fs/open.c | 2 -- include/uapi/linux/openat2.h | 3 +-- tools/testing/selftests/openat2/helpers.h | 3 +-- .../testing/selftests/openat2/openat2_test.c | 24 +++++++------------ 4 files changed, 10 insertions(+), 22 deletions(-)
fs/open.c +0 -2
@@ -993,8 +993,6 @@ static inline int build_open_flags(const struct open_how *how,return -EINVAL ;if (how->resolve & ~VALID_RESOLVE_FLAGS )return -EINVAL ; -if (memchr_inv (how->__padding,0 ,sizeof (how->__padding))) -return -EINVAL ;/* Deal with the mode. */ if (WILL_CREATE (flags)) {
include/uapi/linux/openat2.h +1 -2
@@ -16,8 +16,7 @@*/ struct open_how {__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 mode;__aligned_u64 resolve; };
tools/testing/selftests/openat2/helpers.h +1 -2
@@ -37,8 +37,7 @@*/ struct open_how {__aligned_u64 flags; -__u16 mode; -__u16 __padding[3 ];/* must be zeroed */ +__aligned_u64 mode;__aligned_u64 resolve; };
tools/testing/selftests/openat2/openat2_test.c +8 -16
@@ -40,7 +40,7 @@ struct struct_test {int err; }; -#define NUM_OPENAT2_STRUCT_TESTS 10 +#define NUM_OPENAT2_STRUCT_TESTS 7#define NUM_OPENAT2_STRUCT_VARIATIONS 13void test_openat2_struct (void )
@@ -57,20 +57,6 @@ void test_openat2_struct(void) .arg.inner.flags = O_RDONLY, .size = sizeof(struct open_how_ext) }, - /* Normal struct with broken padding. */ - { .name = "normal struct (non-zero padding[0])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0xa0, 0x00, 0x00}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, - { .name = "normal struct (non-zero padding[1])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0x00, 0x1a, 0x00}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, - { .name = "normal struct (non-zero padding[2])", - .arg.inner.flags = O_RDONLY, - .arg.inner.__padding = {0x00, 0x00, 0xef}, - .size = sizeof(struct open_how_ext), .err = -EINVAL }, -/* TODO: Once expanded, check zero-padding. */ /* Smaller than version-0 struct. */
@@ -169,7 +155,7 @@ struct flag_test {int err; }; -#define NUM_OPENAT2_FLAG_TESTS 21 +#define NUM_OPENAT2_FLAG_TESTS 23void test_openat2_flags (void ) {
@@ -214,9 +200,15 @@ void test_openat2_flags(void) { .name = "invalid how.mode and O_CREAT", .how.flags = O_CREAT, .how.mode = 0xFFFF, .err = -EINVAL }, + { .name = "invalid (very large) how.mode and O_CREAT", + .how.flags = O_CREAT, + .how.mode = 0xC000000000000000ULL, .err = -EINVAL }, { .name = "invalid how.mode and O_TMPFILE", .how.flags = O_TMPFILE | O_RDWR, .how.mode = 0x1337, .err = -EINVAL }, + { .name = "invalid (very large) how.mode and O_TMPFILE", + .how.flags = O_TMPFILE | O_RDWR, + .how.mode = 0x0000A00000000000ULL, .err = -EINVAL }, /* ->resolve must only contain RESOLVE_* flags. */ { .name = "invalid how.resolve and O_RDONLY",
Signature · 1 line
2.24.0
Quoted 8 lines · Aleksa Sarai
While openat2(2) is still not yet in Linus's tree, we can take this opportunity to iron out some small warts that weren't noticed earlier:
* A fix was suggested by Florian Weimer, to separate the openat2 definitions so glibc can use the header directly. I've put the maintainership under VFS but let me know if you'd prefer it belong ot the fcntl folks.
Quoted 6 lines
* Having heterogenous field sizes in an extensible struct results in "padding hole" problems when adding new fields (in addition the correct error to use for non-zero padding isn't entirely clear ). The simplest solution is to just copy clone(3)'s model -- always use u64s. It will waste a little more space in the struct, but it removes a possible future headache.
Am I imagining things or did I get the same patch series twice?
Christian
Quoted 9 lines · Christian Brauner
Quoted 8 lines · Aleksa Sarai
While openat2(2) is still not yet in Linus's tree, we can take this opportunity to iron out some small warts that weren't noticed earlier:
* A fix was suggested by Florian Weimer, to separate the openat2 definitions so glibc can use the header directly. I've put the maintainership under VFS but let me know if you'd prefer it belong ot the fcntl folks.
Quoted 8 lines
Quoted 6 lines
* Having heterogenous field sizes in an extensible struct results in "padding hole" problems when adding new fields (in addition the correct error to use for non-zero padding isn't entirely clear ). The simplest solution is to just copy clone(3)'s model -- always use u64s. It will waste a little more space in the struct, but it removes a possible future headache.
Am I imagining things or did I get the same patch series twice?
Not unless it's a coincidence -- I accidentally ran
% git send-email *.patch [some flags] *.patch
Signature · 4 lines
Aleksa Sarai Senior Software Engineer (Containers) SUSE Linux GmbH <https://www.cyphar.com/>