add file system helpers that take kernel pointers for the init code v4
Subsystems (5)
Hi Al and Linus,
currently a lot of the file system calls in the early in code (and the devtmpfs kthread) rely on the implicit set_fs(KERNEL_DS) during boot. This is one of the few last remaining places we need to deal with to kill off set_fs entirely, so this series adds new helpers that take kernel pointers. These helpers are in init/ and marked __init and thus will be discarded after bootup. A few also need to be duplicated in devtmpfs, though unfortunately.
The series sits on top of my previous
"decruft the early init / initrd / initramfs code v2"
series.
Git tree:
git://git.infradead.org/users/hch/misc.git init_path
Gitweb:
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/init_path
Changes since v3:
- rename fs/for_init.c to fs/init.c
- document the purpose of the routines in fs/init.c with a comment
- don't mark devtmpfs __init as that will cause it to get overwritten by initmem poisoning
- add an init_dup helper to make Al more happy than with the version commit to the "decruft the early init / initrd / initramfs code v2" series
Changes since v2:
- move to fs/for_init.c
- reuse the init routines in devtmpfs after refactoring devtmpfsd (and thus the broken error handling in the previous version)
- actually use kern_path in a place where user_path_at sneaked back in
Changes since v1:
- avoid most core VFS changes
- renamed the functions and move them to init/ and devtmpfs
- drop a bunch of cleanups that can be submitted independently now
Diffstat:
PATCH 1/23 fs: refactor do_mount
Factor out a path_mount helper that takes a struct path * instead of the actual file name. This will allow to convert the init and devtmpfs code to properly mount based on a kernel pointer instead of relying on the implicit set_fs(KERNEL_DS) during early init.
--- fs/namespace.c | 67 ++++++++++++++++++++++++++------------------------ 1 file changed, 35 insertions(+), 32 deletions(-)
fs/namespace.c +35 -32
@@ -3115,12 +3115,11 @@ char *copy_mount_string(const void __user *data)* Therefore, if this magic number is present, it carries no information * and must be discarded. */ -long do_mount (const char *dev_name,const char __user *dir_name, +static int path_mount (const char *dev_name,struct path *path,const char *type_page,unsigned long flags,void *data_page) { -struct path path;unsigned int mnt_flags =0 , sb_flags; -int retval =0 ; +int ret;/* Discard magic */ if ((flags &MS_MGC_MSK ) ==MS_MGC_VAL )
@@ -3133,19 +3132,13 @@ long do_mount(const char *dev_name, const char __user *dir_name,if (flags &MS_NOUSER )return -EINVAL ; -/* ... and get the mountpoint */ - retval =user_path_at (AT_FDCWD , dir_name,LOOKUP_FOLLOW , &path); -if (retval) -return retval; - - retval =security_sb_mount (dev_name, &path, - type_page, flags, data_page); -if (!retval && !may_mount ()) - retval = -EPERM ; -if (!retval && (flags &SB_MANDLOCK ) && !may_mandlock ()) - retval = -EPERM ; -if (retval) - goto dput_out; + ret =security_sb_mount (dev_name, path, type_page, flags, data_page); +if (ret) +return ret; +if (!may_mount ()) +return -EPERM ; +if ((flags &SB_MANDLOCK ) && !may_mandlock ()) +return -EPERM ;/* Default to relatime unless overriden */ if (!(flags &MS_NOATIME ))
@@ -3172,7 +3165,7 @@ long do_mount(const char *dev_name, const char __user *dir_name, ((flags & (MS_NOATIME | MS_NODIRATIME | MS_RELATIME | MS_STRICTATIME)) == 0)) { mnt_flags &= ~MNT_ATIME_MASK ; - mnt_flags |= path.mnt->mnt_flags &MNT_ATIME_MASK ; + mnt_flags |= path->mnt->mnt_flags &MNT_ATIME_MASK ; } sb_flags = flags & (SB_RDONLY |
@@ -3185,22 +3178,32 @@ long do_mount(const char *dev_name, const char __user *dir_name, SB_I_VERSION);if ((flags & (MS_REMOUNT |MS_BIND )) == (MS_REMOUNT |MS_BIND )) - retval =do_reconfigure_mnt (&path, mnt_flags); -else if (flags &MS_REMOUNT ) - retval =do_remount (&path, flags, sb_flags, mnt_flags, - data_page); -else if (flags &MS_BIND ) - retval =do_loopback (&path, dev_name, flags &MS_REC ); -else if (flags & (MS_SHARED |MS_PRIVATE |MS_SLAVE |MS_UNBINDABLE )) - retval =do_change_type (&path, flags); -else if (flags &MS_MOVE ) - retval =do_move_mount_old (&path, dev_name); -else - retval =do_new_mount (&path, type_page, sb_flags, mnt_flags, - dev_name, data_page); -dput_out: +return do_reconfigure_mnt (path, mnt_flags); +if (flags &MS_REMOUNT ) +return do_remount (path, flags, sb_flags, mnt_flags, data_page); +if (flags &MS_BIND ) +return do_loopback (path, dev_name, flags &MS_REC ); +if (flags & (MS_SHARED |MS_PRIVATE |MS_SLAVE |MS_UNBINDABLE )) +return do_change_type (path, flags); +if (flags &MS_MOVE ) +return do_move_mount_old (path, dev_name); + +return do_new_mount (path, type_page, sb_flags, mnt_flags, dev_name, + data_page); +} + +long do_mount (const char *dev_name,const char __user *dir_name, +const char *type_page,unsigned long flags,void *data_page) +{ +struct path path; +int ret; + + ret =user_path_at (AT_FDCWD , dir_name,LOOKUP_FOLLOW , &path); +if (ret) +return ret; + ret =path_mount (dev_name, &path, type_page, flags, data_page);path_put (&path); -return retval; +return ret; }static struct ucounts *inc_mnt_namespaces (struct user_namespace *ns)
Signature · 1 line
2.27.0
PATCH 2/23 fs: refactor ksys_umount
In reply to Christoph HellwigFactor out a path_umount helper that takes a struct path * instead of the actual file name. This will allow to convert the init and devtmpfs code to properly mount based on a kernel pointer instead of relying on the implicit set_fs(KERNEL_DS) during early init.
--- fs/namespace.c | 40 ++++++++++++++++++---------------------- 1 file changed, 18 insertions(+), 22 deletions(-)
fs/namespace.c +18 -22
@@ -1706,36 +1706,19 @@ static inline bool may_mandlock(void) }#endif -/* -* Now umount can handle mount points as well as block devices. -* This is important for filesystems which use unnamed block devices. -* -* We now support a flag for forced unmount like the other 'big iron' -* unixes. Our API is identical to OSF/1 to avoid making a mess of AMD -*/ - -int ksys_umount (char __user *name,int flags) +static int path_umount (struct path *path,int flags) { -struct path path;struct mount *mnt;int retval; -int lookup_flags =LOOKUP_MOUNTPOINT ;if (flags & ~(MNT_FORCE |MNT_DETACH |MNT_EXPIRE |UMOUNT_NOFOLLOW ))return -EINVAL ; -if (!may_mount ())return -EPERM ; -if (!(flags &UMOUNT_NOFOLLOW )) - lookup_flags |=LOOKUP_FOLLOW ; - - retval =user_path_at (AT_FDCWD , name, lookup_flags, &path); -if (retval) - goto out; - mnt =real_mount (path.mnt); + mnt =real_mount (path->mnt); retval = -EINVAL ; -if (path.dentry != path.mnt->mnt_root) +if (path->dentry != path->mnt->mnt_root) goto dput_and_out;if (!check_mnt (mnt)) goto dput_and_out;
@@ -1748,12 +1731,25 @@ int ksys_umount(char __user *name, int flags) retval =do_umount (mnt, flags); dput_and_out:/* we mustn't call path_put() as that would clear mnt_expiry_mark */ -dput (path.dentry); +dput (path->dentry);mntput_no_expire (mnt); -out:return retval; } +int ksys_umount (char __user *name,int flags) +{ +int lookup_flags =LOOKUP_MOUNTPOINT ; +struct path path; +int ret; + +if (!(flags &UMOUNT_NOFOLLOW )) + lookup_flags |=LOOKUP_FOLLOW ; + ret =user_path_at (AT_FDCWD , name, lookup_flags, &path); +if (ret) +return ret; +return path_umount (&path, flags); +} + SYSCALL_DEFINE2(umount, char __user *, name, int, flags) {return ksys_umount (name, flags);
Signature · 1 line
2.27.0
Re: [PATCH 02/23] fs: refactor ksys_umount
Quoted 73 lines · Christoph Hellwig
Factor out a
path_umounthelper that takes a struct path * instead of the actual file name. This will allow to convert the init and devtmpfs code to properly mount based on a kernel pointer instead of relying on the implicitset_fs(KERNEL_DS)during early init.Signed-off-by Christoph Hellwig <[email protected]>--- fs/namespace.c | 40 ++++++++++++++++++---------------------- 1 file changed, 18 insertions(+), 22 deletions(-)fs/namespace.c +11
-22@@ -1706,36 +1706,19 @@ static inline bool may_mandlock(void) }#endif -/* -* Now umount can handle mount points as well as block devices. -* This is important for filesystems which use unnamed block devices. -* -* We now support a flag for forced unmount like the other 'big iron' -* unixes. Our API is identical to OSF/1 to avoid making a mess of AMD -*/ - -int ksys_umount (char __user *name,int flags) +static int path_umount (struct path *path,int flags) { -struct path path;struct mount *mnt;int retval; -int lookup_flags =LOOKUP_MOUNTPOINT ;if (flags & ~(MNT_FORCE |MNT_DETACH |MNT_EXPIRE |UMOUNT_NOFOLLOW ))return -EINVAL ; -if (!may_mount ())return -EPERM ; -if (!(flags &UMOUNT_NOFOLLOW )) - lookup_flags |=LOOKUP_FOLLOW ; - - retval =user_path_at (AT_FDCWD , name, lookup_flags, &path); -if (retval) - goto out; - mnt =real_mount (path.mnt); + mnt =real_mount (path->mnt); retval = -EINVAL ; -if (path.dentry != path.mnt->mnt_root) +if (path->dentry != path->mnt->mnt_root) goto dput_and_out;if (!check_mnt (mnt)) goto dput_and_out;@@ -1748,12 +1731,25 @@ int ksys_umount(char __user *name, int flags) retval =do_umount (mnt, flags); dput_and_out:/* we mustn't call path_put() as that would clear mnt_expiry_mark */ -dput (path.dentry); +dput (path->dentry);mntput_no_expire (mnt); -out:return retval; } +int ksys_umount (char __user *name,int flags) +{ +int lookup_flags =LOOKUP_MOUNTPOINT ; +struct path path; +int ret; + +if (!(flags &UMOUNT_NOFOLLOW ))
fs/namespace.c int ksys_umount(char __user *name, int flags) · line 1755
Quoted 10 lines
fs/namespace.c +7
-0+ lookup_flags |=LOOKUP_FOLLOW ; + ret =user_path_at (AT_FDCWD , name, lookup_flags, &path); +if (ret) +return ret; +return path_umount (&path, flags); +} + SYSCALL_DEFINE2(umount, char __user *, name, int, flags) {return ksys_umount (name, flags);
Regressions on linux next 20200803 tag kernel.
LTP syscalls test umount03 mount a path for testing and umount failed and retired for 50 times and test exit with warning and following test cases using that mount path failed.
LTP syscalls tests failed list,
* umount03
* umount2_01
* umount2_02
* umount2_03
* utime06
* copy_file_range01
Summary
------------------------------------------------------------------------
kernel: 5.8.0
git repo: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
git branch: master
git describe: next-20200803
Test details: https://qa-reports.linaro.org/lkft/linux-next-oe/build/next-20200803
------------------------------------------------------------------------
test failed log:
tst_device.c:262: INFO: Using test device LTP_DEV='/dev/loop0' tst_mkfs.c:90: INFO: Formatting /dev/loop0 with ext2 opts='' extra opts='' mke2fs 1.43.8 (1-Jan-2018)
tst_test.c:1246: INFO: Timeout per run is 0h 15m 00s
[ 870.449934] EXT4-fs (loop0): mounting ext2 file system using the ext4 subsystem [ 870.454338] EXT4-fs (loop0): mounted filesystem without journal. Opts: (null) [ 870.456412] ext2 filesystem being mounted at
/tmp/ltp-YQrzWZNEEy/jVhqum/mntpoint supports timestamps until 2038 (0x7fffffff)
umount03.c:35: PASS: umount() fails as expected: EPERM (1) tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 1... tst_device.c:388: INFO: Likely gvfsd-trash is probing newly mounted fs, kill it to speed up tests.
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 2...
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 3...
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 4...
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 5...
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 6...
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 7...
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 8...
tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 9...tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 10... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 11... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 12... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 13... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 14... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 15... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 16... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 17... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 18... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 19... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 20... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 21... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 22... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 23... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 24... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 25... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 26... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 27... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 28... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 29... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 30... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 31... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 32... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 33... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 34... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 35... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 36... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 37... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 38... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 39... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 40... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 41... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 42... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 43... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 44... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 45... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 46... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 47... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 48... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 49... tst_device.c:384: INFO: umount('mntpoint') failed with EBUSY, try 50... tst_device.c:394: WARN: Failed to umount('mntpoint') after 50 retries tst_tmpdir.c:337: WARN: tst_rmdir: rmobj(/tmp/ltp-YQrzWZNEEy/jVhqum)
failed: remove(/tmp/ltp-YQrzWZNEEy/jVhqum/mntpoint) failed; errno=16:
EBUSY
Summary:
passed 1 failed 0 skipped 0
warnings 1
mke2fs 1.43.8 (1-Jan-2018)/dev/loop0 is mounted; will not make a filesystem here!
umount2_01 0 TINFO : Using test device LTP_DEV='/dev/loop0' umount2_01 0 TINFO : Formatting /dev/loop0 with ext2 opts='' extra opts='' umount2_01 1 TBROK : tst_mkfs.c:103: umount2_01.c:81: mkfs.ext2
failed with 1umount2_01 2 TBROK : tst_mkfs.c:103: Remaining cases broken
mke2fs 1.43.8 (1-Jan-2018)/dev/loop0 is mounted; will not make a filesystem here!
umount2_02 0 TINFO : Using test device LTP_DEV='/dev/loop0' umount2_02 0 TINFO : Formatting /dev/loop0 with ext2 opts='' extra opts='' umount2_02 1 TBROK : tst_mkfs.c:103: umount2_02.c:121: mkfs.ext2
failed with 1umount2_02 2 TBROK : tst_mkfs.c:103: Remaining cases broken
mke2fs 1.43.8 (1-Jan-2018)/dev/loop0 is mounted; will not make a filesystem here!
umount2_03 0 TINFO : Using test device LTP_DEV='/dev/loop0' umount2_03 0 TINFO : Formatting /dev/loop0 with ext2 opts='' extra opts='' umount2_03 1 TBROK : tst_mkfs.c:103: umount2_03.c:101: mkfs.ext2
failed with 1umount2_03 2 TBROK : tst_mkfs.c:103: Remaining cases broken
mke2fs 1.43.8 (1-Jan-2018)/dev/loop0 is mounted; will not make a filesystem here!
utime06 0 TINFO : Using test device LTP_DEV='/dev/loop0' utime06 0 TINFO : Formatting /dev/loop0 with ext2 opts='' extra opts='' utime06 1 TBROK : tst_mkfs.c:103: utime06.c:122: mkfs.ext2 failed with 1 utime06 2 TBROK : tst_mkfs.c:103: Remaining cases broken
Steps to reproduce:
-------------------------
cd /opt/ltp
./runltp -s umount --> FAILS
Above command runs all umount tests.
Test case description,
Verify that umount(2) returns -1 and sets errno to EPERM if the user is not the super-user.
Test case link,
https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/syscalls/umount/umount03.c
full test log,
https://lkft.validation.linaro.org/scheduler/job/1642287
--
Linaro LKFT
https://lkft.linaro.org
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Re: [PATCH 02/23] fs: refactor ksys_umount
Quoted 13 lines · Naresh Kamboju
Quoted 11 lines · Christoph Hellwig
Factor out a
path_umounthelper that takes a struct path * instead of the actual file name. This will allow to convert the init and devtmpfs code to properly mount based on a kernel pointer instead of relying on the implicitset_fs(KERNEL_DS)during early init.Signed-off-by Christoph Hellwig <[email protected]>--- fs/namespace.c | 40 ++++++++++++++++++---------------------- 1 file changed, 18 insertions(+), 22 deletions(-)
Quoted 4 lines
fs/namespace.c
Quoted 4 lines
fs/namespace.c +0
-0
<trim>
Quoted 6 lines
Regressions on linux next 20200803 tag kernel.
LTP syscalls test umount03 mount a path for testing and umount failed and retired for 50 times and test exit with warning and following test cases using that mount path failed.
Quoted 7 lines
LTP syscalls tests failed list,
* umount03
*umount2_01
*umount2_02
*umount2_03
* utime06
*copy_file_range01
The reported issue has been fixed in linux next 20200817 tag by below patch.
fs: fix a struct path leak in path_umount
Make sure we also put the dentry and vfsmnt in the illegal flags and !may_umount cases.
- Naresh
PATCH 3/23 fs: push the getname from do_rmdir into the callers
In reply to Christoph HellwigThis mirrors do_unlinkat and will make life a little easier for the init code to reuse the whole function with a kernel filename.
--- fs/internal.h | 2 +- fs/namei.c | 10 ++++------ include/linux/syscalls.h | 4 ++-- 3 files changed, 7 insertions(+), 9 deletions(-)
fs/internal.h +1 -1
@@ -65,7 +65,7 @@ extern int vfs_path_lookup(struct dentry *, struct vfsmount *,long do_mknodat (int dfd,const char __user *filename,umode_t mode,unsigned int dev);long do_mkdirat (int dfd,const char __user *pathname,umode_t mode); -long do_rmdir (int dfd,const char __user *pathname); +long do_rmdir (int dfd,struct filename *name);long do_unlinkat (int dfd,struct filename *name);long do_symlinkat (const char __user *oldname,int newdfd,const char __user *newname);
fs/namei.c +4 -6
@@ -3720,17 +3720,16 @@ int vfs_rmdir(struct inode *dir, struct dentry *dentry) }EXPORT_SYMBOL (vfs_rmdir); -long do_rmdir (int dfd,const char __user *pathname) +long do_rmdir (int dfd,struct filename *name) {int error =0 ; -struct filename *name;struct dentry *dentry;struct path path;struct qstr last;int type;unsigned int lookup_flags =0 ; retry: - name =filename_parentat (dfd,getname (pathname), lookup_flags, + name =filename_parentat (dfd, name, lookup_flags, &path, &last, &type);if (IS_ERR (name))return PTR_ERR (name);
@@ -3781,7 +3780,7 @@ long do_rmdir(int dfd, const char __user *pathname) SYSCALL_DEFINE1(rmdir, const char __user *, pathname) { -return do_rmdir (AT_FDCWD , pathname); +return do_rmdir (AT_FDCWD ,getname (pathname)); }/**
@@ -3926,8 +3925,7 @@ SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag)return -EINVAL ;if (flag &AT_REMOVEDIR ) -return do_rmdir (dfd, pathname); - +return do_rmdir (dfd,getname (pathname));return do_unlinkat (dfd,getname (pathname)); }
include/linux/syscalls.h +2 -2
@@ -1281,11 +1281,11 @@ static inline long ksys_unlink(const char __user *pathname)return do_unlinkat (AT_FDCWD ,getname (pathname)); } -extern long do_rmdir (int dfd,const char __user *pathname); +long do_rmdir (int dfd,struct filename *name);static inline long ksys_rmdir (const char __user *pathname) { -return do_rmdir (AT_FDCWD , pathname); +return do_rmdir (AT_FDCWD ,getname (pathname)); }extern long do_mkdirat (int dfd,const char __user *pathname,umode_t mode);
Signature · 1 line
2.27.0
PATCH 4/23 devtmpfs: refactor devtmpfsd()
In reply to Christoph HellwigSplit the main worker loop into a separate function. This allows devtmpfsd_setup to be marked __init, which will allows us to call __init routines for the setup work. devtmpfѕ itself needs a __ref marker for that to work, and a comment explaining why it works.
--- drivers/base/devtmpfs.c | 52 ++++++++++++++++++++++++----------------- 1 file changed, 31 insertions(+), 21 deletions(-)
drivers/base/devtmpfs.c +31 -21
@@ -378,7 +378,30 @@ static int handle(const char *name, umode_t mode, kuid_t uid, kgid_t gid,return handle_remove (name, dev); } -static int devtmpfs_setup (void *p) +static void __noreturn devtmpfs_work_loop (void ) +{ +while (1 ) { +spin_lock (&req_lock); +while (requests) { +struct req *req = requests; + requests =NULL ; +spin_unlock (&req_lock); +while (req) { +struct req *next = req->next; + req->err =handle (req->name, req->mode, + req->uid, req->gid, req->dev); +complete (&req->done); + req = next; + } +spin_lock (&req_lock); + } +__set_current_state (TASK_INTERRUPTIBLE ); +spin_unlock (&req_lock); +schedule (); + } +} + +static int __init devtmpfs_setup (void *p) {int err;
@@ -396,31 +419,18 @@ static int devtmpfs_setup(void *p)return err; } -static int devtmpfsd (void *p) +/* +* The __ref is because devtmpfs_setup needs to be __init for the routines it +* calls. That call is done while devtmpfs_init, which is marked __init, +* synchronously waits for it to complete. +*/ +static int __ref devtmpfsd (void *p) {int err =devtmpfs_setup (p);if (err)return err; -while (1 ) { -spin_lock (&req_lock); -while (requests) { -struct req *req = requests; - requests =NULL ; -spin_unlock (&req_lock); -while (req) { -struct req *next = req->next; - req->err =handle (req->name, req->mode, - req->uid, req->gid, req->dev); -complete (&req->done); - req = next; - } -spin_lock (&req_lock); - } -__set_current_state (TASK_INTERRUPTIBLE ); -spin_unlock (&req_lock); -schedule (); - } +devtmpfs_work_loop ();return 0 ; }
Signature · 1 line
2.27.0
Re: [PATCH 04/23] devtmpfs: refactor devtmpfsd()
Quoted 6 lines · Christoph Hellwig
Split the main worker loop into a separate function. This allows
devtmpfsd_setupto be marked__init, which will allows us to call__initroutines for the setup work. devtmpfѕ itself needs a__refmarker for that to work, and a comment explaining why it works.Signed-off-by Christoph Hellwig <[email protected]>
PATCH 5/23 init: initialize ramdisk_execute_command at compile time
In reply to Christoph HellwigSet ramdisk_execute_command to "/init" at compile time. The command line can still override it, but this saves a few instructions and removes a NULL check.
--- init/main.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-)
init/main.c +1 -5
@@ -154,7 +154,7 @@ static bool initargs_found;#endif static char *execute_command; -static char *ramdisk_execute_command; +static char *ramdisk_execute_command ="/init" ;/* * Used to generate warnings if static_key manipulation functions are used
@@ -1514,10 +1514,6 @@ static noinline void __init kernel_init_freeable(void)* check if there is an early userspace init. If yes, let it do all * the work */ - -if (!ramdisk_execute_command) - ramdisk_execute_command ="/init" ; -if (ksys_access ((const char __user *) ramdisk_execute_command,0 ) !=0 ) { ramdisk_execute_command =NULL ;
Signature · 1 line
2.27.0
PATCH 6/23 init: mark console_on_rootfs as __init
In reply to Christoph HellwigThis helper is only used for the early init code.
--- init/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
init/main.c +1 -1
@@ -1458,7 +1458,7 @@ static int __ref kernel_init(void *unused) }/* Open /dev/console, for stdin/stdout/stderr, this should never fail */ -void console_on_rootfs (void ) +void __init console_on_rootfs (void ) {struct file *file =filp_open ("/dev/console" ,O_RDWR ,0 );
Signature · 1 line
2.27.0
PATCH 7/23 init: mark create_dev as __init
In reply to Christoph HellwigThis helper is only used for the early init code.
--- init/do_mounts.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
init/do_mounts.h +1 -1
@@ -13,7 +13,7 @@ void mount_block_root(char *name, int flags);void mount_root (void );extern int root_mountflags; -static inline int create_dev (char *name,dev_t dev) +static inline __init int create_dev (char *name,dev_t dev) {ksys_unlink (name);return ksys_mknod (name,S_IFBLK |0600 ,new_encode_dev (dev));
Signature · 1 line
2.27.0
PATCH 8/23 init: add an init_mount helper
In reply to Christoph HellwigLike do_mount, but takes a kernel pointer for the destination path. Switch over the mounts in the init code and devtmpfs to it, which just happen to work due to the implicit set_fs(KERNEL_DS) during early init right now.
--- drivers/base/devtmpfs.c | 5 +++-- fs/Makefile | 2 +- fs/init.c | 25 +++++++++++++++++++++++++ fs/internal.h | 4 ++++ fs/namespace.c | 2 +- include/linux/init_syscalls.h | 4 ++++ init/do_mounts.c | 8 ++++---- init/do_mounts.h | 1 + init/do_mounts_initrd.c | 6 +++--- 9 files changed, 46 insertions(+), 11 deletions(-) create mode 100644 fs/init.c create mode 100644 include/linux/init_syscalls.h
drivers/base/devtmpfs.c +3 -2
@@ -25,6 +25,7 @@#include <linux/sched.h> #include <linux/slab.h> #include <linux/kthread.h> +#include <linux/init_syscalls.h> #include <uapi/linux/mount.h> #include "base.h"
@@ -359,7 +360,7 @@ int __init devtmpfs_mount(void)if (!thread)return 0 ; - err =do_mount ("devtmpfs" ,"dev" ,"devtmpfs" ,MS_SILENT ,NULL ); + err =init_mount ("devtmpfs" ,"dev" ,"devtmpfs" ,MS_SILENT ,NULL );if (err)printk (KERN_INFO "devtmpfs: error mounting %i\n" , err);else
@@ -408,7 +409,7 @@ static int __init devtmpfs_setup(void *p) err =ksys_unshare (CLONE_NEWNS );if (err) goto out; - err =do_mount ("devtmpfs" ,"/" ,"devtmpfs" ,MS_SILENT ,NULL ); + err =init_mount ("devtmpfs" ,"/" ,"devtmpfs" ,MS_SILENT ,NULL );if (err) goto out;ksys_chdir ("/.." );/* will traverse into overmounted root */
fs/Makefile +1 -1
@@ -13,7 +13,7 @@ obj-y := open.o read_write.o file_table.o super.o \ seq_file.o xattr.o libfs.o fs-writeback.o \ pnode.o splice.o sync.o utimes.o d_path.o \ stack.o fs_struct.o statfs.o fs_pin.o nsfs.o \ - fs_types.o fs_context.o fs_parser.o fsopen.o + fs_types.o fs_context.o fs_parser.o fsopen.o init.o ifeq ($(CONFIG_BLOCK),y) obj-y += buffer.o block_dev.o direct-io.o mpage.o
fs/init.c new file +25 -0
@@ -0,0 +1,25 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Routines that mimic syscalls, but don't use the user address space or file +* descriptors. Only for init/ and related early init code. +*/ +#include <linux/init.h> +#include <linux/mount.h> +#include <linux/namei.h> +#include <linux/fs.h> +#include <linux/init_syscalls.h> +#include "internal.h" + +int __init init_mount (const char *dev_name,const char *dir_name, +const char *type_page,unsigned long flags,void *data_page) +{ +struct path path; +int ret; + + ret =kern_path (dir_name,LOOKUP_FOLLOW , &path); +if (ret) +return ret; + ret =path_mount (dev_name, &path, type_page, flags, data_page); +path_put (&path); +return ret; +}
fs/internal.h +4 -0
@@ -89,6 +89,10 @@ extern int __mnt_want_write_file(struct file *);extern void __mnt_drop_write_file (struct file *);extern void dissolve_on_fput (struct vfsmount *); + +int path_mount (const char *dev_name,struct path *path, +const char *type_page,unsigned long flags,void *data_page); +/* * fs_struct.c */
fs/namespace.c +1 -1
@@ -3111,7 +3111,7 @@ char *copy_mount_string(const void __user *data)* Therefore, if this magic number is present, it carries no information * and must be discarded. */ -static int path_mount (const char *dev_name,struct path *path, +int path_mount (const char *dev_name,struct path *path,const char *type_page,unsigned long flags,void *data_page) {unsigned int mnt_flags =0 , sb_flags;
include/linux/init_syscalls.h new file +4 -0
@@ -0,0 +1,4 @@ +/* SPDX-License-Identifier: GPL-2.0 */ + +int __init init_mount (const char *dev_name,const char *dir_name, +const char *type_page,unsigned long flags,void *data_page);
init/do_mounts.c +4 -4
@@ -395,16 +395,16 @@ static int __init do_mount_root(const char *name, const char *fs,int ret;if (data) { -/* do_mount() requires a full page as fifth argument */ +/* init_mount() requires a full page as fifth argument */ p =alloc_page (GFP_KERNEL );if (!p)return -ENOMEM ; data_page =page_address (p); -/* zero-pad. do_mount() will make sure it's terminated */ +/* zero-pad. init_mount() will make sure it's terminated */ strncpy (data_page, data,PAGE_SIZE ); } - ret =do_mount (name,"/root" , fs, flags, data_page); + ret =init_mount (name,"/root" , fs, flags, data_page);if (ret) goto out;
@@ -628,7 +628,7 @@ void __init prepare_namespace(void)mount_root (); out:devtmpfs_mount (); -do_mount ("." ,"/" ,NULL ,MS_MOVE ,NULL ); +init_mount ("." ,"/" ,NULL ,MS_MOVE ,NULL );ksys_chroot ("." ); }
init/do_mounts.h +1 -0
@@ -8,6 +8,7 @@#include <linux/mount.h> #include <linux/major.h> #include <linux/root_dev.h> +#include <linux/init_syscalls.h> void mount_block_root (char *name,int flags);void mount_root (void );
init/do_mounts_initrd.c +3 -3
@@ -62,7 +62,7 @@ static int __init init_linuxrc(struct subprocess_info *info, struct cred *new)console_on_rootfs ();/* move initrd over / and chdir/chroot in initrd root */ ksys_chdir ("/root" ); -do_mount ("." ,"/" ,NULL ,MS_MOVE ,NULL ); +init_mount ("." ,"/" ,NULL ,MS_MOVE ,NULL );ksys_chroot ("." );ksys_setsid ();return 0 ;
@@ -99,7 +99,7 @@ static void __init handle_initrd(void) current->flags &= ~PF_FREEZER_SKIP ;/* move initrd to rootfs' /old */ -do_mount (".." ,"." ,NULL ,MS_MOVE ,NULL ); +init_mount (".." ,"." ,NULL ,MS_MOVE ,NULL );/* switch root and cwd back to / of rootfs */ ksys_chroot (".." );
@@ -113,7 +113,7 @@ static void __init handle_initrd(void)mount_root ();printk (KERN_NOTICE "Trying to move old root to /initrd ... " ); - error =do_mount ("/old" ,"/root/initrd" ,NULL ,MS_MOVE ,NULL ); + error =init_mount ("/old" ,"/root/initrd" ,NULL ,MS_MOVE ,NULL );if (!error)printk ("okay\n" );else {
Signature · 1 line
2.27.0
PATCH 9/23 init: add an init_umount helper
In reply to Christoph HellwigLike ksys_umount, but takes a kernel pointer for the destination path. Switch over the umount in the init code, which just happen to work due to the implicit set_fs(KERNEL_DS) during early init right now.
--- fs/init.c | 14 ++++++++++++++ fs/internal.h | 1 + fs/namespace.c | 4 ++-- include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 1 - init/do_mounts_initrd.c | 2 +- 6 files changed, 19 insertions(+), 4 deletions(-)
fs/init.c +14 -0
@@ -23,3 +23,17 @@ int __init init_mount(const char *dev_name, const char *dir_name,path_put (&path);return ret; } + +int __init init_umount (const char *name,int flags) +{ +int lookup_flags =LOOKUP_MOUNTPOINT ; +struct path path; +int ret; + +if (!(flags &UMOUNT_NOFOLLOW )) + lookup_flags |=LOOKUP_FOLLOW ; + ret =kern_path (name, lookup_flags, &path); +if (ret) +return ret; +return path_umount (&path, flags); +}
fs/internal.h +1 -0
@@ -92,6 +92,7 @@ extern void dissolve_on_fput(struct vfsmount *);int path_mount (const char *dev_name,struct path *path,const char *type_page,unsigned long flags,void *data_page); +int path_umount (struct path *path,int flags);/* * fs_struct.c
fs/namespace.c +2 -2
@@ -1706,7 +1706,7 @@ static inline bool may_mandlock(void) }#endif -static int path_umount (struct path *path,int flags) +int path_umount (struct path *path,int flags) {struct mount *mnt;int retval;
@@ -1736,7 +1736,7 @@ static int path_umount(struct path *path, int flags)return retval; } -int ksys_umount (char __user *name,int flags) +static int ksys_umount (char __user *name,int flags) {int lookup_flags =LOOKUP_MOUNTPOINT ;struct path path;
include/linux/init_syscalls.h +1 -0
@@ -2,3 +2,4 @@int __init init_mount (const char *dev_name,const char *dir_name,const char *type_page,unsigned long flags,void *data_page); +int __init init_umount (const char *name,int flags);
include/linux/syscalls.h +0 -1
@@ -1236,7 +1236,6 @@ asmlinkage long sys_ni_syscall(void);* the ksys_xyzyyz() functions prototyped below. */ -int ksys_umount (char __user *name,int flags);int ksys_chroot (const char __user *filename);ssize_t ksys_write (unsigned int fd,const char __user *buf,size_t count);int ksys_chdir (const char __user *filename);
init/do_mounts_initrd.c +1 -1
@@ -122,7 +122,7 @@ static void __init handle_initrd(void)else printk ("failed\n" );printk (KERN_NOTICE "Unmounting old root\n" ); -ksys_umount ("/old" ,MNT_DETACH ); +init_umount ("/old" ,MNT_DETACH ); } }
Signature · 1 line
2.27.0
PATCH 10/23 init: add an init_unlink helper
In reply to Christoph HellwigAdd a simple helper to unlink with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_unlink.
--- fs/init.c | 5 +++++ include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 7 ------- init/do_mounts.h | 2 +- init/do_mounts_initrd.c | 4 ++-- init/do_mounts_rd.c | 2 +- init/initramfs.c | 3 ++- 7 files changed, 12 insertions(+), 12 deletions(-)
fs/init.c +5 -0
@@ -37,3 +37,8 @@ int __init init_umount(const char *name, int flags)return ret;return path_umount (&path, flags); } + +int __init init_unlink (const char *pathname) +{ +return do_unlinkat (AT_FDCWD ,getname_kernel (pathname)); +}
include/linux/init_syscalls.h +1 -0
@@ -3,3 +3,4 @@int __init init_mount (const char *dev_name,const char *dir_name,const char *type_page,unsigned long flags,void *data_page);int __init init_umount (const char *name,int flags); +int __init init_unlink (const char *pathname);
include/linux/syscalls.h +0 -7
@@ -1273,13 +1273,6 @@ int compat_ksys_ipc(u32 call, int first, int second,* The following kernel syscall equivalents are just wrappers to fs-internal * functions. Therefore, provide stubs to be inlined at the callsites. */ -extern long do_unlinkat (int dfd,struct filename *name); - -static inline long ksys_unlink (const char __user *pathname) -{ -return do_unlinkat (AT_FDCWD ,getname (pathname)); -} -long do_rmdir (int dfd,struct filename *name);static inline long ksys_rmdir (const char __user *pathname)
init/do_mounts.h +1 -1
@@ -16,7 +16,7 @@ extern int root_mountflags;static inline __init int create_dev (char *name,dev_t dev) { -ksys_unlink (name); +init_unlink (name);return ksys_mknod (name,S_IFBLK |0600 ,new_encode_dev (dev)); }
init/do_mounts_initrd.c +2 -2
@@ -137,11 +137,11 @@ bool __init initrd_load(void)* mounted in the normal path. */ if (rd_load_image ("/initrd.image" ) &&ROOT_DEV != Root_RAM0) { -ksys_unlink ("/initrd.image" ); +init_unlink ("/initrd.image" );handle_initrd ();return true; } } -ksys_unlink ("/initrd.image" ); +init_unlink ("/initrd.image" );return false; }
init/do_mounts_rd.c +1 -1
@@ -272,7 +272,7 @@ int __init rd_load_image(char *from)fput (out_file); out:kfree (buf); -ksys_unlink ("/dev/ram" ); +init_unlink ("/dev/ram" );return res; }
init/initramfs.c +2 -1
@@ -12,6 +12,7 @@#include <linux/file.h> #include <linux/memblock.h> #include <linux/namei.h> +#include <linux/init_syscalls.h> static ssize_t __init xwrite (struct file *file,const char *p,size_t count,loff_t *pos)
@@ -301,7 +302,7 @@ static void __init clean_path(char *path, umode_t fmode)if (S_ISDIR (st.mode))ksys_rmdir (path);else -ksys_unlink (path); +init_unlink (path); } }
Signature · 1 line
2.27.0
PATCH 11/23 init: add an init_rmdir helper
In reply to Christoph HellwigAdd a simple helper to rmdir with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_rmdir.
--- fs/init.c | 5 +++++ include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 7 ------- init/initramfs.c | 2 +- 4 files changed, 7 insertions(+), 8 deletions(-)
fs/init.c +5 -0
@@ -42,3 +42,8 @@ int __init init_unlink(const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname)); } + +int __init init_rmdir (const char *pathname) +{ +return do_rmdir (AT_FDCWD ,getname_kernel (pathname)); +}
include/linux/init_syscalls.h +1 -0
@@ -4,3 +4,4 @@ int __init init_mount(const char *dev_name, const char *dir_name, const char *type_page, unsigned long flags, void *data_page);int __init init_umount (const char *name,int flags);int __init init_unlink (const char *pathname); +int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -7
@@ -1273,13 +1273,6 @@ int compat_ksys_ipc(u32 call, int first, int second,* The following kernel syscall equivalents are just wrappers to fs-internal * functions. Therefore, provide stubs to be inlined at the callsites. */ -long do_rmdir (int dfd,struct filename *name); - -static inline long ksys_rmdir (const char __user *pathname) -{ -return do_rmdir (AT_FDCWD ,getname (pathname)); -} -extern long do_mkdirat (int dfd,const char __user *pathname,umode_t mode);static inline long ksys_mkdir (const char __user *pathname,umode_t mode)
init/initramfs.c +1 -1
@@ -300,7 +300,7 @@ static void __init clean_path(char *path, umode_t fmode)if (!vfs_lstat (path, &st) && (st.mode ^ fmode) &S_IFMT ) {if (S_ISDIR (st.mode)) -ksys_rmdir (path); +init_rmdir (path);else init_unlink (path); }
Signature · 1 line
2.27.0
PATCH 12/23 init: add an init_chdir helper
In reply to Christoph HellwigAdd a simple helper to chdir with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_chdir.
--- drivers/base/devtmpfs.c | 2 +- fs/init.c | 16 ++++++++++++++++ fs/open.c | 7 +------ include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 1 - init/do_mounts.c | 2 +- init/do_mounts_initrd.c | 8 ++++---- 7 files changed, 24 insertions(+), 13 deletions(-)
drivers/base/devtmpfs.c +1 -1
@@ -412,7 +412,7 @@ static int __init devtmpfs_setup(void *p) err =init_mount ("devtmpfs" ,"/" ,"devtmpfs" ,MS_SILENT ,NULL );if (err) goto out; -ksys_chdir ("/.." );/* will traverse into overmounted root */ +init_chdir ("/.." );/* will traverse into overmounted root */ ksys_chroot ("." ); out: *(int *)p = err;
fs/init.c +16 -0
@@ -7,6 +7,7 @@#include <linux/mount.h> #include <linux/namei.h> #include <linux/fs.h> +#include <linux/fs_struct.h> #include <linux/init_syscalls.h> #include "internal.h"
@@ -38,6 +39,21 @@ int __init init_umount(const char *name, int flags)return path_umount (&path, flags); } +int __init init_chdir (const char *filename) +{ +struct path path; +int error; + + error =kern_path (filename,LOOKUP_FOLLOW |LOOKUP_DIRECTORY , &path); +if (error) +return error; + error =inode_permission (path.dentry->d_inode,MAY_EXEC |MAY_CHDIR ); +if (!error) +set_fs_pwd (current->fs, &path); +path_put (&path); +return error; +} +int __init init_unlink (const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname));
fs/open.c +1 -6
@@ -482,7 +482,7 @@ SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)return do_faccessat (AT_FDCWD , filename, mode,0 ); } -int ksys_chdir (const char __user *filename) +SYSCALL_DEFINE1(chdir, const char __user *, filename) {struct path path;int error;
@@ -508,11 +508,6 @@ int ksys_chdir(const char __user *filename)return error; } -SYSCALL_DEFINE1(chdir, const char __user *, filename) -{ -return ksys_chdir (filename); -} - SYSCALL_DEFINE1(fchdir, unsigned int, fd) {struct fd f =fdget_raw (fd);
include/linux/init_syscalls.h +1 -0
@@ -3,5 +3,6 @@int __init init_mount (const char *dev_name,const char *dir_name,const char *type_page,unsigned long flags,void *data_page);int __init init_umount (const char *name,int flags); +int __init init_chdir (const char *filename);int __init init_unlink (const char *pathname);int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -1
@@ -1238,7 +1238,6 @@ asmlinkage long sys_ni_syscall(void);int ksys_chroot (const char __user *filename);ssize_t ksys_write (unsigned int fd,const char __user *buf,size_t count); -int ksys_chdir (const char __user *filename);int ksys_fchown (unsigned int fd,uid_t user,gid_t group);ssize_t ksys_read (unsigned int fd,char __user *buf,size_t count);void ksys_sync (void );
init/do_mounts.c +1 -1
@@ -408,7 +408,7 @@ static int __init do_mount_root(const char *name, const char *fs,if (ret) goto out; -ksys_chdir ("/root" ); +init_chdir ("/root" ); s = current->fs->pwd.dentry->d_sb;ROOT_DEV = s->s_dev; printk(KERN_INFO
init/do_mounts_initrd.c +4 -4
@@ -61,7 +61,7 @@ static int __init init_linuxrc(struct subprocess_info *info, struct cred *new)ksys_unshare (CLONE_FS |CLONE_FILES );console_on_rootfs ();/* move initrd over / and chdir/chroot in initrd root */ -ksys_chdir ("/root" ); +init_chdir ("/root" );init_mount ("." ,"/" ,NULL ,MS_MOVE ,NULL );ksys_chroot ("." );ksys_setsid ();
@@ -82,7 +82,7 @@ static void __init handle_initrd(void)/* mount initrd on rootfs' /root */ mount_block_root ("/dev/root.old" , root_mountflags & ~MS_RDONLY );ksys_mkdir ("/old" ,0700 ); -ksys_chdir ("/old" ); +init_chdir ("/old" );/* * In case that a resume from disk is carried out by linuxrc or one of
@@ -104,11 +104,11 @@ static void __init handle_initrd(void)ksys_chroot (".." );if (new_decode_dev (real_root_dev) == Root_RAM0) { -ksys_chdir ("/old" ); +init_chdir ("/old" );return ; } -ksys_chdir ("/" ); +init_chdir ("/" );ROOT_DEV =new_decode_dev (real_root_dev);mount_root ();
Signature · 1 line
2.27.0
PATCH 13/23 init: add an init_chroot helper
In reply to Christoph HellwigAdd a simple helper to chroot with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_chroot.
--- drivers/base/devtmpfs.c | 2 +- fs/init.c | 24 ++++++++++++++++++++++++ fs/open.c | 7 +------ include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 2 -- init/do_mounts.c | 2 +- init/do_mounts_initrd.c | 4 ++-- 7 files changed, 30 insertions(+), 12 deletions(-)
drivers/base/devtmpfs.c +1 -1
@@ -413,7 +413,7 @@ static int __init devtmpfs_setup(void *p)if (err) goto out;init_chdir ("/.." );/* will traverse into overmounted root */ -ksys_chroot ("." ); +init_chroot ("." ); out: *(int *)p = err;complete (&setup_done);
fs/init.c +24 -0
@@ -9,6 +9,7 @@#include <linux/fs.h> #include <linux/fs_struct.h> #include <linux/init_syscalls.h> +#include <linux/security.h> #include "internal.h" int __init init_mount(const char *dev_name, const char *dir_name,
@@ -54,6 +55,29 @@ int __init init_chdir(const char *filename)return error; } +int __init init_chroot (const char *filename) +{ +struct path path; +int error; + + error =kern_path (filename,LOOKUP_FOLLOW |LOOKUP_DIRECTORY , &path); +if (error) +return error; + error =inode_permission (path.dentry->d_inode,MAY_EXEC |MAY_CHDIR ); +if (error) + goto dput_and_out; + error = -EPERM ; +if (!ns_capable (current_user_ns (),CAP_SYS_CHROOT )) + goto dput_and_out; + error =security_path_chroot (&path); +if (error) + goto dput_and_out; +set_fs_root (current->fs, &path); +dput_and_out: +path_put (&path); +return error; +} +int __init init_unlink (const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname));
fs/open.c +1 -6
@@ -530,7 +530,7 @@ SYSCALL_DEFINE1(fchdir, unsigned int, fd)return error; } -int ksys_chroot (const char __user *filename) +SYSCALL_DEFINE1(chroot, const char __user *, filename) {struct path path;int error;
@@ -563,11 +563,6 @@ int ksys_chroot(const char __user *filename)return error; } -SYSCALL_DEFINE1(chroot, const char __user *, filename) -{ -return ksys_chroot (filename); -} -static int chmod_common (const struct path *path,umode_t mode) {struct inode *inode = path->dentry->d_inode;
include/linux/init_syscalls.h +1 -0
@@ -4,5 +4,6 @@ int __init init_mount(const char *dev_name, const char *dir_name, const char *type_page, unsigned long flags, void *data_page);int __init init_umount (const char *name,int flags);int __init init_chdir (const char *filename); +int __init init_chroot (const char *filename);int __init init_unlink (const char *pathname);int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -2
@@ -1235,8 +1235,6 @@ asmlinkage long sys_ni_syscall(void);* Instead, use one of the functions which work equivalently, such as * the ksys_xyzyyz() functions prototyped below. */ - -int ksys_chroot (const char __user *filename);ssize_t ksys_write (unsigned int fd,const char __user *buf,size_t count);int ksys_fchown (unsigned int fd,uid_t user,gid_t group);ssize_t ksys_read (unsigned int fd,char __user *buf,size_t count);
init/do_mounts.c +1 -1
@@ -629,7 +629,7 @@ void __init prepare_namespace(void) out:devtmpfs_mount ();init_mount ("." ,"/" ,NULL ,MS_MOVE ,NULL ); -ksys_chroot ("." ); +init_chroot ("." ); }static bool is_tmpfs;
init/do_mounts_initrd.c +2 -2
@@ -63,7 +63,7 @@ static int __init init_linuxrc(struct subprocess_info *info, struct cred *new)/* move initrd over / and chdir/chroot in initrd root */ init_chdir ("/root" );init_mount ("." ,"/" ,NULL ,MS_MOVE ,NULL ); -ksys_chroot ("." ); +init_chroot ("." );ksys_setsid ();return 0 ; }
@@ -101,7 +101,7 @@ static void __init handle_initrd(void)/* move initrd to rootfs' /old */ init_mount (".." ,"." ,NULL ,MS_MOVE ,NULL );/* switch root and cwd back to / of rootfs */ -ksys_chroot (".." ); +init_chroot (".." );if (new_decode_dev (real_root_dev) == Root_RAM0) {init_chdir ("/old" );
Signature · 1 line
2.27.0
PATCH 14/23 init: add an init_chown helper
In reply to Christoph HellwigAdd a simple helper to chown with a kernel space file name and switch the early init code over to it.
--- fs/init.c | 18 ++++++++++++++++++ fs/internal.h | 2 +- fs/open.c | 2 +- include/linux/init_syscalls.h | 1 + init/initramfs.c | 6 +++--- 5 files changed, 24 insertions(+), 5 deletions(-)
fs/init.c +18 -0
@@ -78,6 +78,24 @@ int __init init_chroot(const char *filename)return error; } +int __init init_chown (const char *filename,uid_t user,gid_t group,int flags) +{ +int lookup_flags = (flags &AT_SYMLINK_NOFOLLOW ) ?0 :LOOKUP_FOLLOW ; +struct path path; +int error; + + error =kern_path (filename, lookup_flags, &path); +if (error) +return error; + error =mnt_want_write (path.mnt); +if (!error) { + error =chown_common (&path, user, group); +mnt_drop_write (path.mnt); + } +path_put (&path); +return error; +} +int __init init_unlink (const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname));
fs/internal.h +1 -1
@@ -134,7 +134,7 @@ long do_sys_ftruncate(unsigned int fd, loff_t length, int small);int do_fchmodat (int dfd,const char __user *filename,umode_t mode);int do_fchownat (int dfd,const char __user *filename,uid_t user,gid_t group,int flag); - +int chown_common (const struct path *path,uid_t user,gid_t group);extern int vfs_open (const struct path *,struct file *);/*
fs/open.c +1 -1
@@ -639,7 +639,7 @@ SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)return do_fchmodat (AT_FDCWD , filename, mode); } -static int chown_common (const struct path *path,uid_t user,gid_t group) +int chown_common (const struct path *path,uid_t user,gid_t group) {struct inode *inode = path->dentry->d_inode;struct inode *delegated_inode =NULL ;
include/linux/init_syscalls.h +1 -0
@@ -5,5 +5,6 @@ int __init init_mount(const char *dev_name, const char *dir_name,int __init init_umount (const char *name,int flags);int __init init_chdir (const char *filename);int __init init_chroot (const char *filename); +int __init init_chown (const char *filename,uid_t user,gid_t group,int flags);int __init init_unlink (const char *pathname);int __init init_rmdir (const char *pathname);
init/initramfs.c +3 -3
@@ -349,14 +349,14 @@ static int __init do_name(void) } }else if (S_ISDIR (mode)) {ksys_mkdir (collected, mode); -ksys_chown (collected, uid, gid); +init_chown (collected, uid, gid,0 );ksys_chmod (collected, mode);dir_add (collected, mtime); }else if (S_ISBLK (mode) ||S_ISCHR (mode) ||S_ISFIFO (mode) ||S_ISSOCK (mode)) {if (maybe_link () ==0 ) {ksys_mknod (collected, mode, rdev); -ksys_chown (collected, uid, gid); +init_chown (collected, uid, gid,0 );ksys_chmod (collected, mode);do_utime (collected, mtime); }
@@ -394,7 +394,7 @@ static int __init do_symlink(void) collected[N_ALIGN (name_len) + body_len] ='\0' ;clean_path (collected,0 );ksys_symlink (collected +N_ALIGN (name_len), collected); -ksys_lchown (collected, uid, gid); +init_chown (collected, uid, gid,AT_SYMLINK_NOFOLLOW );do_utime (collected, mtime); state = SkipIt; next_state = Reset;
Signature · 1 line
2.27.0
PATCH 15/23 init: add an init_chmod helper
In reply to Christoph HellwigAdd a simple helper to chmod with a kernel space file name and switch the early init code over to it.
--- fs/init.c | 13 +++++++++++++ fs/internal.h | 2 +- fs/open.c | 4 ++-- include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 7 ------- init/initramfs.c | 4 ++-- 6 files changed, 19 insertions(+), 12 deletions(-)
fs/init.c +13 -0
@@ -96,6 +96,19 @@ int __init init_chown(const char *filename, uid_t user, gid_t group, int flags)return error; } +int __init init_chmod (const char *filename,umode_t mode) +{ +struct path path; +int error; + + error =kern_path (filename,LOOKUP_FOLLOW , &path); +if (error) +return error; + error =chmod_common (&path, mode); +path_put (&path); +return error; +} +int __init init_unlink (const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname));
fs/internal.h +1 -1
@@ -131,7 +131,7 @@ extern struct open_how build_open_how(int flags, umode_t mode);extern int build_open_flags (const struct open_how *how,struct open_flags *op);long do_sys_ftruncate (unsigned int fd,loff_t length,int small); -int do_fchmodat (int dfd,const char __user *filename,umode_t mode); +int chmod_common (const struct path *path,umode_t mode);int do_fchownat (int dfd,const char __user *filename,uid_t user,gid_t group,int flag);int chown_common (const struct path *path,uid_t user,gid_t group);
fs/open.c +2 -2
@@ -563,7 +563,7 @@ SYSCALL_DEFINE1(chroot, const char __user *, filename)return error; } -static int chmod_common (const struct path *path,umode_t mode) +int chmod_common (const struct path *path,umode_t mode) {struct inode *inode = path->dentry->d_inode;struct inode *delegated_inode =NULL ;
@@ -610,7 +610,7 @@ SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)return err; } -int do_fchmodat (int dfd,const char __user *filename,umode_t mode) +static int do_fchmodat (int dfd,const char __user *filename,umode_t mode) {struct path path;int error;
include/linux/init_syscalls.h +1 -0
@@ -6,5 +6,6 @@ int __init init_umount(const char *name, int flags);int __init init_chdir (const char *filename);int __init init_chroot (const char *filename);int __init init_chown (const char *filename,uid_t user,gid_t group,int flags); +int __init init_chmod (const char *filename,umode_t mode);int __init init_unlink (const char *pathname);int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -7
@@ -1304,13 +1304,6 @@ static inline long ksys_link(const char __user *oldname,return do_linkat (AT_FDCWD , oldname,AT_FDCWD , newname,0 ); } -extern int do_fchmodat (int dfd,const char __user *filename,umode_t mode); - -static inline int ksys_chmod (const char __user *filename,umode_t mode) -{ -return do_fchmodat (AT_FDCWD , filename, mode); -} -long do_faccessat (int dfd,const char __user *filename,int mode,int flags);static inline long ksys_access (const char __user *filename,int mode)
init/initramfs.c +2 -2
@@ -350,14 +350,14 @@ static int __init do_name(void) }else if (S_ISDIR (mode)) {ksys_mkdir (collected, mode);init_chown (collected, uid, gid,0 ); -ksys_chmod (collected, mode); +init_chmod (collected, mode);dir_add (collected, mtime); }else if (S_ISBLK (mode) ||S_ISCHR (mode) ||S_ISFIFO (mode) ||S_ISSOCK (mode)) {if (maybe_link () ==0 ) {ksys_mknod (collected, mode, rdev);init_chown (collected, uid, gid,0 ); -ksys_chmod (collected, mode); +init_chmod (collected, mode);do_utime (collected, mtime); } }
Signature · 1 line
2.27.0
PATCH 16/23 init: add an init_eaccess helper
In reply to Christoph HellwigAdd a simple helper to check if a file exists based on kernel space file name and switch the early init code over to it. Note that this theoretically changes behavior as it always is based on the effective permissions. But during early init that doesn't make a difference.
--- fs/init.c | 13 +++++++++++++ fs/open.c | 2 +- include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 7 ------- init/main.c | 4 ++-- 5 files changed, 17 insertions(+), 10 deletions(-)
fs/init.c +13 -0
@@ -109,6 +109,19 @@ int __init init_chmod(const char *filename, umode_t mode)return error; } +int __init init_eaccess (const char *filename) +{ +struct path path; +int error; + + error =kern_path (filename,LOOKUP_FOLLOW , &path); +if (error) +return error; + error =inode_permission (d_inode (path.dentry),MAY_ACCESS ); +path_put (&path); +return error; +} +int __init init_unlink (const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname));
fs/open.c +1 -1
@@ -394,7 +394,7 @@ static const struct cred *access_override_creds(void)return old_cred; } -long do_faccessat (int dfd,const char __user *filename,int mode,int flags) +static long do_faccessat (int dfd,const char __user *filename,int mode,int flags) {struct path path;struct inode *inode;
include/linux/init_syscalls.h +1 -0
@@ -7,5 +7,6 @@ int __init init_chdir(const char *filename);int __init init_chroot (const char *filename);int __init init_chown (const char *filename,uid_t user,gid_t group,int flags);int __init init_chmod (const char *filename,umode_t mode); +int __init init_eaccess (const char *filename);int __init init_unlink (const char *pathname);int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -7
@@ -1304,13 +1304,6 @@ static inline long ksys_link(const char __user *oldname,return do_linkat (AT_FDCWD , oldname,AT_FDCWD , newname,0 ); } -long do_faccessat (int dfd,const char __user *filename,int mode,int flags); - -static inline long ksys_access (const char __user *filename,int mode) -{ -return do_faccessat (AT_FDCWD , filename, mode,0 ); -} -extern int do_fchownat (int dfd,const char __user *filename,uid_t user,gid_t group,int flag);
init/main.c +2 -2
@@ -96,6 +96,7 @@#include <linux/jump_label.h> #include <linux/mem_encrypt.h> #include <linux/kcsan.h> +#include <linux/init_syscalls.h> #include <asm/io.h> #include <asm/bugs.h>
@@ -1514,8 +1515,7 @@ static noinline void __init kernel_init_freeable(void)* check if there is an early userspace init. If yes, let it do all * the work */ -if (ksys_access ((const char __user *) - ramdisk_execute_command,0 ) !=0 ) { +if (init_eaccess (ramdisk_execute_command) !=0 ) { ramdisk_execute_command =NULL ;prepare_namespace (); }
Signature · 1 line
2.27.0
PATCH 17/23 init: add an init_link helper
In reply to Christoph HellwigAdd a simple helper to link with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_link.
--- fs/init.c | 33 +++++++++++++++++++++++++++++++++ fs/internal.h | 3 +-- fs/namei.c | 4 ++-- include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 9 --------- init/initramfs.c | 2 +- 6 files changed, 38 insertions(+), 14 deletions(-)
fs/init.c +33 -0
@@ -122,6 +122,39 @@ int __init init_eaccess(const char *filename)return error; } +int __init init_link (const char *oldname,const char *newname) +{ +struct dentry *new_dentry; +struct path old_path, new_path; +int error; + + error =kern_path (oldname,0 , &old_path); +if (error) +return error; + + new_dentry =kern_path_create (AT_FDCWD , newname, &new_path,0 ); + error =PTR_ERR (new_dentry); +if (IS_ERR (new_dentry)) + goto out; + + error = -EXDEV ; +if (old_path.mnt != new_path.mnt) + goto out_dput; + error =may_linkat (&old_path); +if (unlikely (error)) + goto out_dput; + error =security_path_link (old_path.dentry, &new_path, new_dentry); +if (error) + goto out_dput; + error =vfs_link (old_path.dentry, new_path.dentry->d_inode, new_dentry, +NULL ); +out_dput: +done_path_create (&new_path, new_dentry); +out: +path_put (&old_path); +return error; +} +int __init init_unlink (const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname));
fs/internal.h +1 -2
@@ -69,8 +69,7 @@ long do_rmdir(int dfd, struct filename *name);long do_unlinkat (int dfd,struct filename *name);long do_symlinkat (const char __user *oldname,int newdfd,const char __user *newname); -int do_linkat (int olddfd,const char __user *oldname,int newdfd, -const char __user *newname,int flags); +int may_linkat (struct path *link);/* * namespace.c
fs/namei.c +2 -2
@@ -1024,7 +1024,7 @@ static bool safe_hardlink_source(struct inode *inode)* * Returns 0 if successful, -ve on error. */ -static int may_linkat (struct path *link) +int may_linkat (struct path *link) {struct inode *inode = link->dentry->d_inode;
@@ -4086,7 +4086,7 @@ EXPORT_SYMBOL(vfs_link);* with linux 2.0, and to avoid hard-linking to directories * and other special files. --ADM */ -int do_linkat (int olddfd,const char __user *oldname,int newdfd, +static int do_linkat (int olddfd,const char __user *oldname,int newdfd,const char __user *newname,int flags) {struct dentry *new_dentry;
include/linux/init_syscalls.h +1 -0
@@ -8,5 +8,6 @@ int __init init_chroot(const char *filename);int __init init_chown (const char *filename,uid_t user,gid_t group,int flags);int __init init_chmod (const char *filename,umode_t mode);int __init init_eaccess (const char *filename); +int __init init_link (const char *oldname,const char *newname);int __init init_unlink (const char *pathname);int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -9
@@ -1295,15 +1295,6 @@ static inline long ksys_mknod(const char __user *filename, umode_t mode,return do_mknodat (AT_FDCWD , filename, mode, dev); } -extern int do_linkat (int olddfd,const char __user *oldname,int newdfd, -const char __user *newname,int flags); - -static inline long ksys_link (const char __user *oldname, -const char __user *newname) -{ -return do_linkat (AT_FDCWD , oldname,AT_FDCWD , newname,0 ); -} -extern int do_fchownat (int dfd,const char __user *filename,uid_t user,gid_t group,int flag);
init/initramfs.c +1 -1
@@ -312,7 +312,7 @@ static int __init maybe_link(void)char *old =find_link (major, minor, ino, mode, collected);if (old) {clean_path (collected,0 ); -return (ksys_link (old, collected) <0 ) ?-1 :1 ; +return (init_link (old, collected) <0 ) ?-1 :1 ; } }return 0 ;
Signature · 1 line
2.27.0
PATCH 18/23 init: add an init_symlink helper
In reply to Christoph HellwigAdd a simple helper to symlink with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_symlink.
--- fs/init.c | 16 ++++++++++++++++ fs/internal.h | 2 -- fs/namei.c | 2 +- include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 9 --------- init/initramfs.c | 2 +- 6 files changed, 19 insertions(+), 13 deletions(-)
fs/init.c +16 -0
@@ -155,6 +155,22 @@ int __init init_link(const char *oldname, const char *newname)return error; } +int __init init_symlink (const char *oldname,const char *newname) +{ +struct dentry *dentry; +struct path path; +int error; + + dentry =kern_path_create (AT_FDCWD , newname, &path,0 ); +if (IS_ERR (dentry)) +return PTR_ERR (dentry); + error =security_path_symlink (&path, dentry, oldname); +if (!error) + error =vfs_symlink (path.dentry->d_inode, dentry, oldname); +done_path_create (&path, dentry); +return error; +} +int __init init_unlink (const char *pathname) {return do_unlinkat (AT_FDCWD ,getname_kernel (pathname));
fs/internal.h +0 -2
@@ -67,8 +67,6 @@ long do_mknodat(int dfd, const char __user *filename, umode_t mode,long do_mkdirat (int dfd,const char __user *pathname,umode_t mode);long do_rmdir (int dfd,struct filename *name);long do_unlinkat (int dfd,struct filename *name); -long do_symlinkat (const char __user *oldname,int newdfd, -const char __user *newname);int may_linkat (struct path *link);/*
fs/namei.c +1 -1
@@ -3955,7 +3955,7 @@ int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) }EXPORT_SYMBOL (vfs_symlink); -long do_symlinkat (const char __user *oldname,int newdfd, +static long do_symlinkat (const char __user *oldname,int newdfd,const char __user *newname) {int error;
include/linux/init_syscalls.h +1 -0
@@ -9,5 +9,6 @@ int __init init_chown(const char *filename, uid_t user, gid_t group, int flags);int __init init_chmod (const char *filename,umode_t mode);int __init init_eaccess (const char *filename);int __init init_link (const char *oldname,const char *newname); +int __init init_symlink (const char *oldname,const char *newname);int __init init_unlink (const char *pathname);int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -9
@@ -1277,15 +1277,6 @@ static inline long ksys_mkdir(const char __user *pathname, umode_t mode)return do_mkdirat (AT_FDCWD , pathname, mode); } -extern long do_symlinkat (const char __user *oldname,int newdfd, -const char __user *newname); - -static inline long ksys_symlink (const char __user *oldname, -const char __user *newname) -{ -return do_symlinkat (oldname,AT_FDCWD , newname); -} -extern long do_mknodat (int dfd,const char __user *filename,umode_t mode,unsigned int dev);
init/initramfs.c +1 -1
@@ -393,7 +393,7 @@ static int __init do_symlink(void) { collected[N_ALIGN (name_len) + body_len] ='\0' ;clean_path (collected,0 ); -ksys_symlink (collected +N_ALIGN (name_len), collected); +init_symlink (collected +N_ALIGN (name_len), collected);init_chown (collected, uid, gid,AT_SYMLINK_NOFOLLOW );do_utime (collected, mtime); state = SkipIt;
Signature · 1 line
2.27.0
PATCH 19/23 init: add an init_mkdir helper
In reply to Christoph HellwigAdd a simple helper to mkdir with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_mkdir.
--- fs/init.c | 18 ++++++++++++++++++ fs/internal.h | 1 - fs/namei.c | 2 +- include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 7 ------- init/do_mounts_initrd.c | 2 +- init/initramfs.c | 2 +- init/noinitramfs.c | 5 +++-- 8 files changed, 25 insertions(+), 13 deletions(-)
fs/init.c +18 -0
@@ -176,6 +176,24 @@ int __init init_unlink(const char *pathname)return do_unlinkat (AT_FDCWD ,getname_kernel (pathname)); } +int __init init_mkdir (const char *pathname,umode_t mode) +{ +struct dentry *dentry; +struct path path; +int error; + + dentry =kern_path_create (AT_FDCWD , pathname, &path,LOOKUP_DIRECTORY ); +if (IS_ERR (dentry)) +return PTR_ERR (dentry); +if (!IS_POSIXACL (path.dentry->d_inode)) + mode &= ~current_umask (); + error =security_path_mkdir (&path, dentry, mode); +if (!error) + error =vfs_mkdir (path.dentry->d_inode, dentry, mode); +done_path_create (&path, dentry); +return error; +} +int __init init_rmdir (const char *pathname) {return do_rmdir (AT_FDCWD ,getname_kernel (pathname));
fs/internal.h +0 -1
@@ -64,7 +64,6 @@ extern int vfs_path_lookup(struct dentry *, struct vfsmount *, const char *, unsigned int, struct path *);long do_mknodat (int dfd,const char __user *filename,umode_t mode,unsigned int dev); -long do_mkdirat (int dfd,const char __user *pathname,umode_t mode);long do_rmdir (int dfd,struct filename *name);long do_unlinkat (int dfd,struct filename *name);int may_linkat (struct path *link);
fs/namei.c +1 -1
@@ -3645,7 +3645,7 @@ int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) }EXPORT_SYMBOL (vfs_mkdir); -long do_mkdirat (int dfd,const char __user *pathname,umode_t mode) +static long do_mkdirat (int dfd,const char __user *pathname,umode_t mode) {struct dentry *dentry;struct path path;
include/linux/init_syscalls.h +1 -0
@@ -11,4 +11,5 @@ int __init init_eaccess(const char *filename);int __init init_link (const char *oldname,const char *newname);int __init init_symlink (const char *oldname,const char *newname);int __init init_unlink (const char *pathname); +int __init init_mkdir (const char *pathname,umode_t mode);int __init init_rmdir (const char *pathname);
include/linux/syscalls.h +0 -7
@@ -1270,13 +1270,6 @@ int compat_ksys_ipc(u32 call, int first, int second,* The following kernel syscall equivalents are just wrappers to fs-internal * functions. Therefore, provide stubs to be inlined at the callsites. */ -extern long do_mkdirat (int dfd,const char __user *pathname,umode_t mode); - -static inline long ksys_mkdir (const char __user *pathname,umode_t mode) -{ -return do_mkdirat (AT_FDCWD , pathname, mode); -} -extern long do_mknodat (int dfd,const char __user *filename,umode_t mode,unsigned int dev);
init/do_mounts_initrd.c +1 -1
@@ -81,7 +81,7 @@ static void __init handle_initrd(void)create_dev ("/dev/root.old" , Root_RAM0);/* mount initrd on rootfs' /root */ mount_block_root ("/dev/root.old" , root_mountflags & ~MS_RDONLY ); -ksys_mkdir ("/old" ,0700 ); +init_mkdir ("/old" ,0700 );init_chdir ("/old" );/*
init/initramfs.c +1 -1
@@ -348,7 +348,7 @@ static int __init do_name(void) state = CopyFile; } }else if (S_ISDIR (mode)) { -ksys_mkdir (collected, mode); +init_mkdir (collected, mode);init_chown (collected, uid, gid,0 );init_chmod (collected, mode);dir_add (collected, mtime);
init/noinitramfs.c +3 -2
@@ -9,6 +9,7 @@#include <linux/stat.h> #include <linux/kdev_t.h> #include <linux/syscalls.h> +#include <linux/init_syscalls.h> /* * Create a simple rootfs that is similar to the default initramfs
@@ -17,7 +18,7 @@ static int __init default_rootfs(void) {int err; - err =ksys_mkdir ((const char __user __force *)"/dev" ,0755 ); + err =init_mkdir ("/dev" ,0755 );if (err <0 ) goto out;
@@ -27,7 +28,7 @@ static int __init default_rootfs(void)if (err <0 ) goto out; - err =ksys_mkdir ((const char __user __force *)"/root" ,0700 ); + err =init_mkdir ("/root" ,0700 );if (err <0 ) goto out;
Signature · 1 line
2.27.0
PATCH 20/23 init: add an init_mknod helper
In reply to Christoph HellwigAdd a simple helper to mknod with a kernel space file name and switch the early init code over to it. Remove the now unused ksys_mknod.
--- fs/init.c | 25 +++++++++++++++++++++++++ fs/internal.h | 2 -- fs/namei.c | 2 +- include/linux/init_syscalls.h | 1 + include/linux/syscalls.h | 9 --------- init/do_mounts.h | 2 +- init/initramfs.c | 2 +- init/noinitramfs.c | 3 +-- 8 files changed, 30 insertions(+), 16 deletions(-)
fs/init.c +25 -0
@@ -122,6 +122,31 @@ int __init init_eaccess(const char *filename)return error; } +int __init init_mknod (const char *filename,umode_t mode,unsigned int dev) +{ +struct dentry *dentry; +struct path path; +int error; + +if (S_ISFIFO (mode) ||S_ISSOCK (mode)) + dev =0 ; +else if (!(S_ISBLK (mode) ||S_ISCHR (mode))) +return -EINVAL ; + + dentry =kern_path_create (AT_FDCWD , filename, &path,0 ); +if (IS_ERR (dentry)) +return PTR_ERR (dentry); + +if (!IS_POSIXACL (path.dentry->d_inode)) + mode &= ~current_umask (); + error =security_path_mknod (&path, dentry, mode, dev); +if (!error) + error =vfs_mknod (path.dentry->d_inode, dentry, mode, +new_decode_dev (dev)); +done_path_create (&path, dentry); +return error; +} +int __init init_link (const char *oldname,const char *newname) {struct dentry *new_dentry;
fs/internal.h +0 -2
@@ -62,8 +62,6 @@ extern int filename_lookup(int dfd, struct filename *name, unsigned flags, struct path *path, struct path *root);extern int vfs_path_lookup (struct dentry *,struct vfsmount *,const char *,unsigned int ,struct path *); -long do_mknodat (int dfd,const char __user *filename,umode_t mode, -unsigned int dev);long do_rmdir (int dfd,struct filename *name);long do_unlinkat (int dfd,struct filename *name);int may_linkat (struct path *link);
fs/namei.c +1 -1
@@ -3564,7 +3564,7 @@ static int may_mknod(umode_t mode) } } -long do_mknodat (int dfd,const char __user *filename,umode_t mode, +static long do_mknodat (int dfd,const char __user *filename,umode_t mode,unsigned int dev) {struct dentry *dentry;
include/linux/init_syscalls.h +1 -0
@@ -8,6 +8,7 @@ int __init init_chroot(const char *filename);int __init init_chown (const char *filename,uid_t user,gid_t group,int flags);int __init init_chmod (const char *filename,umode_t mode);int __init init_eaccess (const char *filename); +int __init init_mknod (const char *filename,umode_t mode,unsigned int dev);int __init init_link (const char *oldname,const char *newname);int __init init_symlink (const char *oldname,const char *newname);int __init init_unlink (const char *pathname);
include/linux/syscalls.h +0 -9
@@ -1270,15 +1270,6 @@ int compat_ksys_ipc(u32 call, int first, int second,* The following kernel syscall equivalents are just wrappers to fs-internal * functions. Therefore, provide stubs to be inlined at the callsites. */ -extern long do_mknodat (int dfd,const char __user *filename,umode_t mode, -unsigned int dev); - -static inline long ksys_mknod (const char __user *filename,umode_t mode, -unsigned int dev) -{ -return do_mknodat (AT_FDCWD , filename, mode, dev); -} -extern int do_fchownat (int dfd,const char __user *filename,uid_t user,gid_t group,int flag);
init/do_mounts.h +1 -1
@@ -17,7 +17,7 @@ extern int root_mountflags;static inline __init int create_dev (char *name,dev_t dev) {init_unlink (name); -return ksys_mknod (name,S_IFBLK |0600 ,new_encode_dev (dev)); +return init_mknod (name,S_IFBLK |0600 ,new_encode_dev (dev)); }#ifdef CONFIG_BLK_DEV_RAM
init/initramfs.c +1 -1
@@ -355,7 +355,7 @@ static int __init do_name(void) }else if (S_ISBLK (mode) ||S_ISCHR (mode) ||S_ISFIFO (mode) ||S_ISSOCK (mode)) {if (maybe_link () ==0 ) { -ksys_mknod (collected, mode, rdev); +init_mknod (collected, mode, rdev);init_chown (collected, uid, gid,0 );init_chmod (collected, mode);do_utime (collected, mtime);
init/noinitramfs.c +1 -2
@@ -22,8 +22,7 @@ static int __init default_rootfs(void)if (err <0 ) goto out; - err =ksys_mknod ((const char __user __force *)"/dev/console" , -S_IFCHR |S_IRUSR |S_IWUSR , + err =init_mknod ("/dev/console" ,S_IFCHR |S_IRUSR |S_IWUSR ,new_encode_dev (MKDEV (5 ,1 )));if (err <0 ) goto out;
Signature · 1 line
2.27.0
PATCH 21/23 init: add an init_stat helper
In reply to Christoph HellwigAdd a simple helper to stat with a kernel space file name and switch the early init code over to it.
--- drivers/md/md-autodetect.c | 3 ++- fs/init.c | 15 +++++++++++++++ include/linux/init_syscalls.h | 1 + init/initramfs.c | 3 ++- 4 files changed, 20 insertions(+), 2 deletions(-)
drivers/md/md-autodetect.c +2 -1
@@ -5,6 +5,7 @@#include <linux/mount.h> #include <linux/major.h> #include <linux/delay.h> +#include <linux/init_syscalls.h> #include <linux/raid/detect.h> #include <linux/raid/md_u.h> #include <linux/raid/md_p.h>
@@ -151,7 +152,7 @@ static void __init md_setup_drive(struct md_setup_args *args)if (strncmp (devname,"/dev/" ,5 ) ==0 ) devname +=5 ;snprintf (comp_name,63 ,"/dev/%s" , devname); -if (vfs_stat (comp_name, &stat) ==0 &&S_ISBLK (stat.mode)) +if (init_stat (comp_name, &stat,0 ) ==0 &&S_ISBLK (stat.mode)) dev =new_decode_dev (stat.rdev);if (!dev) {pr_warn ("md: Unknown device name: %s\n" , devname);
fs/init.c +15 -0
@@ -122,6 +122,21 @@ int __init init_eaccess(const char *filename)return error; } +int __init init_stat (const char *filename,struct kstat *stat,int flags) +{ +int lookup_flags = (flags &AT_SYMLINK_NOFOLLOW ) ?0 :LOOKUP_FOLLOW ; +struct path path; +int error; + + error =kern_path (filename, lookup_flags, &path); +if (error) +return error; + error =vfs_getattr (&path, stat,STATX_BASIC_STATS , + flags |AT_NO_AUTOMOUNT ); +path_put (&path); +return error; +} +int __init init_mknod (const char *filename,umode_t mode,unsigned int dev) {struct dentry *dentry;
include/linux/init_syscalls.h +1 -0
@@ -8,6 +8,7 @@ int __init init_chroot(const char *filename);int __init init_chown (const char *filename,uid_t user,gid_t group,int flags);int __init init_chmod (const char *filename,umode_t mode);int __init init_eaccess (const char *filename); +int __init init_stat (const char *filename,struct kstat *stat,int flags);int __init init_mknod (const char *filename,umode_t mode,unsigned int dev);int __init init_link (const char *oldname,const char *newname);int __init init_symlink (const char *oldname,const char *newname);
init/initramfs.c +2 -1
@@ -298,7 +298,8 @@ static void __init clean_path(char *path, umode_t fmode) {struct kstat st; -if (!vfs_lstat (path, &st) && (st.mode ^ fmode) &S_IFMT ) { +if (init_stat (path, &st,AT_SYMLINK_NOFOLLOW ) && + (st.mode ^ fmode) &S_IFMT ) {if (S_ISDIR (st.mode))init_rmdir (path);else
Signature · 1 line
2.27.0
PATCH 1/1 init: fix error check in clean_path()
init_stat() returns 0 on success, same as vfs_lstat(). When it replaced vfs_lstat(), the '!' was dropped.
---
Andy: this was broken in virtme. "/init: source: not found"
init/initramfs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
init/initramfs.c +1 -1
@@ -297,7 +297,7 @@ static void __init clean_path(char *path, umode_t fmode) {struct kstat st; -if (init_stat (path, &st,AT_SYMLINK_NOFOLLOW ) && +if (!init_stat (path, &st,AT_SYMLINK_NOFOLLOW ) && (st.mode ^ fmode) &S_IFMT ) {if (S_ISDIR (st.mode))init_rmdir (path);
Signature · 1 line
2.28.0.526.ge36021eeef-goog
Re: [PATCH] init: fix error check in clean_path()
Quoted 5 lines · Barret Rhoden
init_stat()returns 0 on success, same asvfs_lstat(). When it replacedvfs_lstat(), the '!' was dropped.Fixes 716308a5331b ("init: add an init_stat helper")Signed-off-by Barret Rhoden <[email protected]>
Looks good,
PATCH 22/23 init: add an init_utimes helper
In reply to Christoph HellwigAdd a simple helper to set timestamps with a kernel space file name and switch the early init code over to it.
--- fs/init.c | 13 +++++++++++++ include/linux/init_syscalls.h | 1 + init/initramfs.c | 3 +-- 3 files changed, 15 insertions(+), 2 deletions(-)
fs/init.c +13 -0
@@ -238,3 +238,16 @@ int __init init_rmdir(const char *pathname) {return do_rmdir (AT_FDCWD ,getname_kernel (pathname)); } + +int __init init_utimes (char *filename,struct timespec64 *ts) +{ +struct path path; +int error; + + error =kern_path (filename,0 , &path); +if (error) +return error; + error =vfs_utimes (&path, ts); +path_put (&path); +return error; +}
include/linux/init_syscalls.h +1 -0
@@ -15,3 +15,4 @@ int __init init_symlink(const char *oldname, const char *newname);int __init init_unlink (const char *pathname);int __init init_mkdir (const char *pathname,umode_t mode);int __init init_rmdir (const char *pathname); +int __init init_utimes (char *filename,struct timespec64 *ts);
init/initramfs.c +1 -2
@@ -111,8 +111,7 @@ static long __init do_utime(char *filename, time64_t mtime) t[0 ].tv_nsec =0 ; t[1 ].tv_sec = mtime; t[1 ].tv_nsec =0 ; - -return do_utimes (AT_FDCWD , filename, t,AT_SYMLINK_NOFOLLOW ); +return init_utimes (filename, t); }static __initdata LIST_HEAD (dir_list);
Signature · 1 line
2.27.0
PATCH 23/23 init: add an init_dup helper
In reply to Christoph HellwigAdd a simple helper to grab a reference to a file and install it at the next available fd, and switch the early init code over to it.
--- fs/init.c | 12 ++++++++++++ include/linux/init_syscalls.h | 1 + init/main.c | 7 +++---- 3 files changed, 16 insertions(+), 4 deletions(-)
fs/init.c +12 -0
@@ -8,6 +8,7 @@#include <linux/namei.h> #include <linux/fs.h> #include <linux/fs_struct.h> +#include <linux/file.h> #include <linux/init_syscalls.h> #include <linux/security.h> #include "internal.h"
@@ -251,3 +252,14 @@ int __init init_utimes(char *filename, struct timespec64 *ts)path_put (&path);return error; } + +int __init init_dup (struct file *file) +{ +int fd; + + fd =get_unused_fd_flags (0 ); +if (fd <0 ) +return fd; +fd_install (get_unused_fd_flags (0 ),get_file (file)); +return 0 ; +}
include/linux/init_syscalls.h +1 -0
@@ -16,3 +16,4 @@ int __init init_unlink(const char *pathname);int __init init_mkdir (const char *pathname,umode_t mode);int __init init_rmdir (const char *pathname);int __init init_utimes (char *filename,struct timespec64 *ts); +int __init init_dup (struct file *file);
init/main.c +3 -4
@@ -1467,10 +1467,9 @@ void __init console_on_rootfs(void)pr_err ("Warning: unable to open an initial console.\n" );return ; } -get_file_rcu_many (file,2 ); -fd_install (get_unused_fd_flags (0 ), file); -fd_install (get_unused_fd_flags (0 ), file); -fd_install (get_unused_fd_flags (0 ), file); +init_dup (file); +init_dup (file); +init_dup (file); } static noinline void __init kernel_init_freeable(void)
Signature · 1 line
2.27.0
Al Viro
maintainerQuoted 10 lines · Christoph Hellwig
Hi Al and Linus,
currently a lot of the file system calls in the early in code (and the devtmpfs kthread) rely on the implicit
set_fs(KERNEL_DS)during boot. This is one of the few last remaining places we need to deal with to kill offset_fsentirely, so this series adds new helpers that take kernel pointers. These helpers are ininit/and marked__initand thus will be discarded after bootup. A few also need to be duplicated in devtmpfs, though unfortunately.
Quoted 3 lines
The series sits on top of my previous
"decruft the early init / initrd / initramfs code v2"
Could you fold the fixes in the parent branch to avoid the bisect hazards? As it is, you have e.g. "initd: pass a non-f_pos offset to kernel_read/kernel_write" that ought to go into "initrd: switch initrd loading to struct file based APIs"...
Quoted 11 lines · Al Viro
Quoted 10 lines · Christoph Hellwig
Hi Al and Linus,
currently a lot of the file system calls in the early in code (and the devtmpfs kthread) rely on the implicit
set_fs(KERNEL_DS)during boot. This is one of the few last remaining places we need to deal with to kill offset_fsentirely, so this series adds new helpers that take kernel pointers. These helpers are ininit/and marked__initand thus will be discarded after bootup. A few also need to be duplicated in devtmpfs, though unfortunately.
Quoted 7 lines
Quoted 3 lines
The series sits on top of my previous
"decruft the early init / initrd / initramfs code v2"
Could you fold the fixes in the parent branch to avoid the bisect hazards? As it is, you have e.g. "initd: pass a non-f_pos offset to kernel_read/kernel_write" that ought to go into "initrd: switch initrd loading to struct file based APIs"...
I'm not a huge fan of rebasing after it has been out for a long time and with pending other patches on top of it. But at your request I've now folded the fixes and force pushed it.
Al Viro
maintainerQuoted 16 lines · Christoph Hellwig
Quoted 15 lines · Al Viro
Quoted 13 lines · Christoph Hellwig
Hi Al and Linus,
currently a lot of the file system calls in the early in code (and the devtmpfs kthread) rely on the implicit
set_fs(KERNEL_DS)during boot. This is one of the few last remaining places we need to deal with to kill offset_fsentirely, so this series adds new helpers that take kernel pointers. These helpers are ininit/and marked__initand thus will be discarded after bootup. A few also need to be duplicated in devtmpfs, though unfortunately.The series sits on top of my previous
"decruft the early init / initrd / initramfs code v2"
Quoted 7 lines
Quoted 3 lines
Could you fold the fixes in the parent branch to avoid the bisect hazards? As it is, you have e.g. "initd: pass a non-f_pos offset to kernel_read/kernel_write" that ought to go into "initrd: switch initrd loading to struct file based APIs"...
I'm not a huge fan of rebasing after it has been out for a long time and with pending other patches on top of it. But at your request I've now folded the fixes and force pushed it.
Um...
Christoph Hellwig (28):
[snip]
initramfs: switch initramfs unpacking to struct file based APIs
initramfs: switch initramfs unpacking to struct file based APIs
[snip]
It's not a bisect hazard, of course, but if you don't fold those together, you might at least want to give the second one a different commit summary... I hadn't been able to find an analogue of #init_path on top of that either.
As it is, #init-user-pointers is fine (aside of that SNAFU with unfolded pair of commits), and so's the contents of #init_path part following what used to be #init-user-pointers, but it'll be an awful mess on merge in the current shape.
I can sort it out myself, if you don't mind that; again, I'm OK with the contents and I've no problem with doing reordering/folding.
Quoted 11 lines · Al Viro
Christoph Hellwig (28):
[snip]
initramfs: switch initramfs unpacking to struct file based APIs
initramfs: switch initramfs unpacking to struct file based APIs
[snip]It's not a bisect hazard, of course, but if you don't fold those together, you might at least want to give the second one a different commit summary... I hadn't been able to find an analogue of #init_path on top of that either.
Quoted 7 lines
As it is, #init-user-pointers is fine (aside of that SNAFU with unfolded pair of commits), and so's the contents of #init_path part following what used to be #init-user-pointers, but it'll be an awful mess on merge in the current shape.
I can sort it out myself, if you don't mind that; again, I'm OK with the contents and I've no problem with doing reordering/folding.
I've fixed the folding issues in init-user-pointers and rebased init_path on top of that. Feel free to pull it. I don't think it is worth reposting.
Quoted 9 lines · Christoph Hellwig
Hi Al and Linus,
currently a lot of the file system calls in the early in code (and the devtmpfs kthread) rely on the implicit
set_fs(KERNEL_DS)during boot. This is one of the few last remaining places we need to deal with to kill offset_fsentirely, so this series adds new helpers that take kernel pointers. These helpers are ininit/and marked__initand thus will be discarded after bootup. A few also need to be duplicated in devtmpfs, though unfortunately.
Reverting this series from next-20200803 fixed the crash below on shutdown.
[ 7303.287890][ T1] systemd-shutdown[1]: All loop devices detached. [ 7303.287930][ T1] systemd-shutdown[1]: Detaching DM devices. [ 7303.441674][ T1] printk: shutdown: 9 output lines suppressed due to ratelimiting [ 7303.443999][ T1] Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000100 [ 7303.444027][ T1] CPU: 120 PID: 1 Comm: shutdown Not tainted 5.8.0-next-20200803 #2 [ 7303.444053][ T1] Call Trace: [ 7303.444069][ T1] [c000000015d27b70] [c0000000006f3778] dump_stack+0xfc/0x174 (unreliable) [ 7303.444103][ T1] [c000000015d27bc0] [c0000000000c9328] panic+0x214/0x4ac [ 7303.444140][ T1] [c000000015d27c60] [c0000000000d4d28] do_exit+0xda8/0xee0 [ 7303.444178][ T1] [c000000015d27d60] [c0000000000d4f1c] do_group_exit+0x5c/0xd0 list_empty at include/linux/list.h:282 (inlined by) thread_group_empty at include/linux/sched/signal.h:671 (inlined by) do_group_exit at kernel/exit.c:888 [ 7303.444205][ T1] [c000000015d27da0] [c0000000000d4fac] sys_exit_group+0x1c/0x20 sys_exit_group at kernel/exit.c:914 [ 7303.444234][ T1] [c000000015d27dc0] [c00000000002c628] system_call_exception+0xf8/0x1d0 [ 7303.444262][ T1] [c000000015d27e20] [c00000000000d0a8] system_call_common+0xe8/0x218 [ 7304.936912][ T1] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000100 ]---
Quoted 39 lines
The series sits on top of my previous
"decruft the early init / initrd / initramfs code v2"
series.
Git tree:
git://git.infradead.org/users/hch/misc.gitinit_pathGitweb:
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/init_path
Changes since v3:
- renamefs/for_init.ctofs/init.c
- document the purpose of the routines infs/init.cwith a comment
- don't mark devtmpfs__initas that will cause it to get overwritten by initmem poisoning
- add aninit_duphelper to make Al more happy than with the version commit to the "decruft the early init / initrd / initramfs code v2" seriesChanges since v2:
- move tofs/for_init.c
- reuse the init routines in devtmpfs after refactoring devtmpfsd (and thus the broken error handling in the previous version)
- actually usekern_pathin a place whereuser_path_atsneaked back inChanges since v1:
- avoid most core VFS changes
- renamed the functions and move them toinit/and devtmpfs
- drop a bunch of cleanups that can be submitted independently nowDiffstat:
Quoted 11 lines · Qian Cai
Quoted 9 lines · Christoph Hellwig
Hi Al and Linus,
currently a lot of the file system calls in the early in code (and the devtmpfs kthread) rely on the implicit
set_fs(KERNEL_DS)during boot. This is one of the few last remaining places we need to deal with to kill offset_fsentirely, so this series adds new helpers that take kernel pointers. These helpers are ininit/and marked__initand thus will be discarded after bootup. A few also need to be duplicated in devtmpfs, though unfortunately.
Quoted 1 line
Reverting this series from next-20200803 fixed the crash below on shutdown.
Please try this patch:
---
From 6448eebe2fe7189cedc5136ab3464517956922b7 Mon Sep 17 00:00:00 2001
From: Christoph Hellwig <[email protected]>
Date: Mon, 3 Aug 2020 15:56:18 +0200
Subject: init: fix init_dup
Don't allocate an unused fd for each call. Also drop the extra reference from filp_open after the init_dup calls while we're at it.
Reported-by Stephen Rothwell <[email protected]>
--- fs/init.c | 2 +- init/main.c | 1 + 2 files changed, 2 insertions(+), 1 deletion(-)
fs/init.c +1 -1
@@ -260,6 +260,6 @@ int __init init_dup(struct file *file) fd =get_unused_fd_flags (0 );if (fd <0 )return fd; -fd_install (get_unused_fd_flags (0 ),get_file (file)); +fd_install (fd,get_file (file));return 0 ; }
init/main.c +1 -0
@@ -1470,6 +1470,7 @@ void __init console_on_rootfs(void)init_dup (file);init_dup (file);init_dup (file); +fput (file); } static noinline void __init kernel_init_freeable(void)
Signature · 1 line
2.27.0