kernel.git/fs/fcntl.c, branch linux-6.10.y

fs: Fix file_set_fowner LSM hook inconsistencies

2024-10-04T14:33:41Z

commit 26f204380a3c182e5adf1a798db0724d6111b597 upstream. The fcntl's F_SETOWN command sets the process that handle SIGIO/SIGURG for the related file descriptor. Before this change, the file_set_fowner LSM hook was always called, ignoring the VFS logic which may not actually change the process that handles SIGIO (e.g. TUN, TTY, dnotify), nor update the related UID/EUID. Moreover, because security_file_set_fowner() was called without lock (e.g. f_owner.lock), concurrent F_SETOWN commands could result to a race condition and inconsistent LSM states (e.g. SELinux's fown_sid) compared to struct fown_struct's UID/EUID. This change makes sure the LSM states are always in sync with the VFS state by moving the security_file_set_fowner() call close to the UID/EUID updates and using the same f_owner.lock . Rename f_modown() to __f_setown() to simplify code. Cc: stable@vger.kernel.org Cc: Al Viro Cc: Casey Schaufler Cc: Christian Brauner Cc: James Morris Cc: Jann Horn Cc: Ondrej Mosnacek Cc: Paul Moore Cc: Serge E. Hallyn Cc: Stephen Smalley Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Mickaël Salaün Signed-off-by: Paul Moore Signed-off-by: Greg Kroah-Hartman

fcntl: add F_DUPFD_QUERY fcntl()

2024-05-10T06:26:31Z

Often userspace needs to know whether two file descriptors refer to the same struct file. For example, systemd uses this to filter out duplicate file descriptors in it's file descriptor store (cf. [1]) and vulkan uses it to compare dma-buf fds (cf. [2]). The only api we provided for this was kcmp() but that's not generally available or might be disallowed because it is way more powerful (allows ordering of file pointers, operates on non-current task) etc. So give userspace a simple way of comparing two file descriptors for sameness adding a new fcntl() F_DUDFD_QUERY. Link: https://github.com/systemd/systemd/blob/a4f0e0da3573a10bc5404142be8799418760b1d1/src/basic/fd-util.c#L517 [1] Link: https://gitlab.freedesktop.org/wlroots/wlroots/-/blob/master/render/vulkan/texture.c#L490 [2] Signed-off-by: Linus Torvalds [brauner: commit message] Signed-off-by: Christian Brauner

Merge tag 'vfs-6.9.iomap' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

2024-03-11T17:07:03Z

Pull iomap updates from Christian Brauner: - Restore read-write hints in struct bio through the bi_write_hint member for the sake of UFS devices in mobile applications. This can result in up to 40% lower write amplification in UFS devices. The patch series that builds on this will be coming in via the SCSI maintainers (Bart) - Overhaul the iomap writeback code. Afterwards ->map_blocks() is able to map multiple blocks at once as long as they're in the same folio. This reduces CPU usage for buffered write workloads on e.g., xfs on systems with lots of cores (Christoph) - Record processed bytes in iomap_iter() trace event (Kassey) - Extend iomap_writepage_map() trace event after Christoph's ->map_block() changes to map mutliple blocks at once (Zhang) * tag 'vfs-6.9.iomap' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (22 commits) iomap: Add processed for iomap_iter iomap: add pos and dirty_len into trace_iomap_writepage_map block, fs: Restore the per-bio/request data lifetime fields fs: Propagate write hints to the struct block_device inode fs: Move enum rw_hint into a new header file fs: Split fcntl_rw_hint() fs: Verify write lifetime constants at compile time fs: Fix rw_hint validation iomap: pass the length of the dirty region to ->map_blocks iomap: map multiple blocks at a time iomap: submit ioends immediately iomap: factor out a iomap_writepage_map_block helper iomap: only call mapping_set_error once for each failed bio iomap: don't chain bios iomap: move the iomap_sector sector calculation out of iomap_add_to_ioend iomap: clean up the iomap_alloc_ioend calling convention iomap: move all remaining per-folio logic into iomap_writepage_map iomap: factor out a iomap_writepage_handle_eof helper iomap: move the PF_MEMALLOC check to iomap_writepages iomap: move the io_folios field out of struct iomap_ioend ...

fs: prefer kfree_rcu() in fasync_remove_entry()

2024-02-12T10:00:10Z

In 'fasync_remove_entry()', prefer 'kfree_rcu()' over 'call_rcu()' with dummy 'fasync_free_rcu()' callback. This is mostly intended in attempt to fix weird https://syzkaller.appspot.com/bug?id=6a64ad907e361e49e92d1c4c114128a1bda2ed7f, where kmemleak may consider 'fa' as unreferenced during RCU grace period. See https://lore.kernel.org/stable/20230930174657.800551-1-joel@joelfernandes.org as well. Comments are highly appreciated. Ever since ae65a5211d90 ("mm/slab: document kfree() as allowed for kmem_cache_alloc() objects") kfree() can be used for both kmalloc() and kmem_cache_alloc() so this is no safe. Do not backport this to stable, please. Link ae65a5211d90 ("mm/slab: document kfree() as > allowed for kmem_cache_alloc() objects") Signed-off-by: Dmitry Antipov Link: https://lore.kernel.org/r/20240209125220.330383-1-dmantipov@yandex.ru Signed-off-by: Christian Brauner

fs: Propagate write hints to the struct block_device inode

2024-02-06T13:30:48Z

Write hints applied with F_SET_RW_HINT on a block device affect the block device inode only. Propagate these hints to the inode associated with struct block_device because that is the inode used when writing back dirty pages. Cc: Alexander Viro Cc: Christian Brauner Cc: Jens Axboe Cc: Christoph Hellwig Cc: Jeff Layton Cc: Chuck Lever Cc: Kanchan Joshi Signed-off-by: Bart Van Assche Link: https://lore.kernel.org/r/20240202203926.2478590-6-bvanassche@acm.org Signed-off-by: Christian Brauner

fs: Move enum rw_hint into a new header file

2024-02-06T13:30:48Z

Move enum rw_hint into a new header file to prepare for using this data type in the block layer. Add the attribute __packed to reduce the space occupied by instances of this data type from four bytes to one byte. Change the data type of i_write_hint from u8 into enum rw_hint. Reviewed-by: Christoph Hellwig Acked-by: Chao Yu # for the F2FS part Cc: Alexander Viro Cc: Christian Brauner Cc: Jan Kara Cc: Christoph Hellwig Signed-off-by: Bart Van Assche Link: https://lore.kernel.org/r/20240202203926.2478590-5-bvanassche@acm.org Signed-off-by: Christian Brauner

fs: Split fcntl_rw_hint()

2024-02-06T13:30:47Z

Split fcntl_rw_hint() such that there is one helper function per fcntl. Use READ_ONCE() and WRITE_ONCE() to access the i_write_hint member instead of protecting such accesses with the inode lock. READ_ONCE() is not used in I/O path code that reads i_write_hint. Users who want F_SET_RW_HINT to affect I/O need to make sure that F_SET_RW_HINT has completed before I/O is submitted that should use the configured write hint. Cc: Christoph Hellwig Suggested-by: Christoph Hellwig Cc: Kanchan Joshi Cc: Jeff Layton Cc: Chuck Lever Cc: Jens Axboe Cc: Stephen Rothwell Signed-off-by: Bart Van Assche Link: https://lore.kernel.org/r/20240202203926.2478590-4-bvanassche@acm.org Signed-off-by: Christian Brauner

fs: Verify write lifetime constants at compile time

2024-02-06T13:30:47Z

The code in fs/fcntl.c converts RWH_* constants to and from WRITE_LIFE_* constants using casts. Verify at compile time that these casts will yield the intended effect. Reviewed-by: Christoph Hellwig Suggested-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Signed-off-by: Bart Van Assche Link: https://lore.kernel.org/r/20240202203926.2478590-3-bvanassche@acm.org Signed-off-by: Christian Brauner

fs: Fix rw_hint validation

2024-02-06T13:30:47Z

Reject values that are valid rw_hints after truncation but not before truncation by passing an untruncated value to rw_hint_valid(). Reviewed-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Cc: Jeff Layton Cc: Chuck Lever Cc: Jens Axboe Cc: Stephen Rothwell Fixes: 5657cb0797c4 ("fs/fcntl: use copy_to/from_user() for u64 types") Signed-off-by: Bart Van Assche Link: https://lore.kernel.org/r/20240202203926.2478590-2-bvanassche@acm.org Signed-off-by: Christian Brauner

treewide: mark stuff as __ro_after_init

2023-10-18T21:43:23Z

__read_mostly predates __ro_after_init. Many variables which are marked __read_mostly should have been __ro_after_init from day 1. Also, mark some stuff as "const" and "__init" while I'm at it. [akpm@linux-foundation.org: revert sysctl_nr_open_min, sysctl_nr_open_max changes due to arm warning] [akpm@linux-foundation.org: coding-style cleanups] Link: https://lkml.kernel.org/r/4f6bb9c0-abba-4ee4-a7aa-89265e886817@p183 Signed-off-by: Alexey Dobriyan Signed-off-by: Andrew Morton