From e7478158e1378325907edfdd960eca98a1be405b Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Wed, 29 Jun 2022 15:06:57 +0200 Subject: fs: clear or set FMODE_LSEEK based on llseek function Pipe-like behaviour on llseek(2) (i.e. unconditionally failing with -ESPIPE) can be expresses in 3 ways: 1) ->llseek set to NULL in file_operations 2) ->llseek set to no_llseek in file_operations 3) FMODE_LSEEK *not* set in ->f_mode. Enforce (3) in cases (1) and (2); that will allow to simplify the checks and eventually get rid of no_llseek boilerplate. Signed-off-by: Jason A. Donenfeld Signed-off-by: Al Viro --- fs/file_table.c | 2 ++ fs/open.c | 4 ++++ 2 files changed, 6 insertions(+) (limited to 'fs') diff --git a/fs/file_table.c b/fs/file_table.c index 5424e3a8df5f..0658b822beeb 100644 --- a/fs/file_table.c +++ b/fs/file_table.c @@ -235,6 +235,8 @@ static struct file *alloc_file(const struct path *path, int flags, file->f_mapping = path->dentry->d_inode->i_mapping; file->f_wb_err = filemap_sample_wb_err(file->f_mapping); file->f_sb_err = file_sample_sb_err(file); + if (fop->llseek && fop->llseek != no_llseek) + file->f_mode |= FMODE_LSEEK; if ((file->f_mode & FMODE_READ) && likely(fop->read || fop->read_iter)) file->f_mode |= FMODE_CAN_READ; diff --git a/fs/open.c b/fs/open.c index 1d57fbde2feb..4488bd77c390 100644 --- a/fs/open.c +++ b/fs/open.c @@ -858,6 +858,10 @@ static int do_dentry_open(struct file *f, if ((f->f_mode & FMODE_WRITE) && likely(f->f_op->write || f->f_op->write_iter)) f->f_mode |= FMODE_CAN_WRITE; + if ((f->f_mode & FMODE_LSEEK) && !f->f_op->llseek) + f->f_mode &= ~FMODE_LSEEK; + if ((f->f_mode & FMODE_LSEEK) && f->f_op->llseek == no_llseek) + f->f_mode &= ~FMODE_LSEEK; if (f->f_mapping->a_ops && f->f_mapping->a_ops->direct_IO) f->f_mode |= FMODE_CAN_ODIRECT; -- cgit v1.2.3 From 4e3299eaddffd9d7d5b8bae28ad700bb775f02d0 Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Wed, 29 Jun 2022 15:06:59 +0200 Subject: fs: do not compare against ->llseek Now vfs_llseek() can simply check for FMODE_LSEEK; if it's set, we know that ->llseek() won't be NULL and if it's not we should just fail with -ESPIPE. A couple of other places where we used to check for special values of ->llseek() (somewhat inconsistently) switched to checking FMODE_LSEEK. Signed-off-by: Jason A. Donenfeld Signed-off-by: Al Viro --- fs/coredump.c | 4 ++-- fs/overlayfs/copy_up.c | 3 +-- fs/read_write.c | 11 +++-------- 3 files changed, 6 insertions(+), 12 deletions(-) (limited to 'fs') diff --git a/fs/coredump.c b/fs/coredump.c index ebc43f960b64..9f4aae202109 100644 --- a/fs/coredump.c +++ b/fs/coredump.c @@ -816,9 +816,9 @@ static int __dump_skip(struct coredump_params *cprm, size_t nr) { static char zeroes[PAGE_SIZE]; struct file *file = cprm->file; - if (file->f_op->llseek && file->f_op->llseek != no_llseek) { + if (file->f_mode & FMODE_LSEEK) { if (dump_interrupted() || - file->f_op->llseek(file, nr, SEEK_CUR) < 0) + vfs_llseek(file, nr, SEEK_CUR) < 0) return 0; cprm->pos += nr; return 1; diff --git a/fs/overlayfs/copy_up.c b/fs/overlayfs/copy_up.c index 714ec569d25b..022606c9703b 100644 --- a/fs/overlayfs/copy_up.c +++ b/fs/overlayfs/copy_up.c @@ -226,8 +226,7 @@ static int ovl_copy_up_data(struct ovl_fs *ofs, struct path *old, /* Couldn't clone, so now we try to copy the data */ /* Check if lower fs supports seek operation */ - if (old_file->f_mode & FMODE_LSEEK && - old_file->f_op->llseek) + if (old_file->f_mode & FMODE_LSEEK) skip_hole = true; while (len) { diff --git a/fs/read_write.c b/fs/read_write.c index b1b1cdfee9d3..d94b6dbba6f9 100644 --- a/fs/read_write.c +++ b/fs/read_write.c @@ -290,14 +290,9 @@ EXPORT_SYMBOL(default_llseek); loff_t vfs_llseek(struct file *file, loff_t offset, int whence) { - loff_t (*fn)(struct file *, loff_t, int); - - fn = no_llseek; - if (file->f_mode & FMODE_LSEEK) { - if (file->f_op->llseek) - fn = file->f_op->llseek; - } - return fn(file, offset, whence); + if (!(file->f_mode & FMODE_LSEEK)) + return -ESPIPE; + return file->f_op->llseek(file, offset, whence); } EXPORT_SYMBOL(vfs_llseek); -- cgit v1.2.3 From 97ef77c52b789ec1411d360ed99dca1efe4b2c81 Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Wed, 29 Jun 2022 15:06:58 +0200 Subject: fs: check FMODE_LSEEK to control internal pipe splicing The original direct splicing mechanism from Jens required the input to be a regular file because it was avoiding the special socket case. It also recognized blkdevs as being close enough to a regular file. But it forgot about chardevs, which behave the same way and work fine here. This is an okayish heuristic, but it doesn't totally work. For example, a few chardevs should be spliceable here. And a few regular files shouldn't. This patch fixes this by instead checking whether FMODE_LSEEK is set, which represents decently enough what we need rewinding for when splicing to internal pipes. Fixes: b92ce5589374 ("[PATCH] splice: add direct fd <-> fd splicing support") Cc: Jens Axboe Signed-off-by: Jason A. Donenfeld Signed-off-by: Al Viro --- fs/splice.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/splice.c b/fs/splice.c index 047b79db8eb5..93a2c9bf6249 100644 --- a/fs/splice.c +++ b/fs/splice.c @@ -814,17 +814,15 @@ ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd, { struct pipe_inode_info *pipe; long ret, bytes; - umode_t i_mode; size_t len; int i, flags, more; /* - * We require the input being a regular file, as we don't want to - * randomly drop data for eg socket -> socket splicing. Use the - * piped splicing for that! + * We require the input to be seekable, as we don't want to randomly + * drop data for eg socket -> socket splicing. Use the piped splicing + * for that! */ - i_mode = file_inode(in)->i_mode; - if (unlikely(!S_ISREG(i_mode) && !S_ISBLK(i_mode))) + if (unlikely(!(in->f_mode & FMODE_LSEEK))) return -EINVAL; /* -- cgit v1.2.3 From 868941b14441282ba08761b770fc6cad69d5bdb7 Mon Sep 17 00:00:00 2001 From: "Jason A. Donenfeld" Date: Wed, 29 Jun 2022 15:07:00 +0200 Subject: fs: remove no_llseek Now that all callers of ->llseek are going through vfs_llseek(), we don't gain anything by keeping no_llseek around. Nothing actually calls it and setting ->llseek to no_lseek is completely equivalent to leaving it NULL. Longer term (== by the end of merge window) we want to remove all such intializations. To simplify the merge window this commit does *not* touch initializers - it only defines no_llseek as NULL (and simplifies the tests on file opening). At -rc1 we'll need do a mechanical removal of no_llseek - git grep -l -w no_llseek | grep -v porting.rst | while read i; do sed -i '/\/d' $i done would do it. Signed-off-by: Jason A. Donenfeld Signed-off-by: Al Viro --- Documentation/filesystems/porting.rst | 8 ++++++++ drivers/gpu/drm/drm_file.c | 3 +-- fs/file_table.c | 2 +- fs/open.c | 2 -- fs/read_write.c | 6 ------ include/linux/fs.h | 2 +- kernel/bpf/bpf_iter.c | 3 +-- 7 files changed, 12 insertions(+), 14 deletions(-) (limited to 'fs') diff --git a/Documentation/filesystems/porting.rst b/Documentation/filesystems/porting.rst index 2e0e4f0e0c6f..aee9aaf9f3df 100644 --- a/Documentation/filesystems/porting.rst +++ b/Documentation/filesystems/porting.rst @@ -914,3 +914,11 @@ Calling conventions for file_open_root() changed; now it takes struct path * instead of passing mount and dentry separately. For callers that used to pass mnt_root> pair (i.e. the root of given mount), a new helper is provided - file_open_root_mnt(). In-tree users adjusted. + +--- + +**mandatory** + +no_llseek is gone; don't set .llseek to that - just leave it NULL instead. +Checks for "does that file have llseek(2), or should it fail with ESPIPE" +should be done by looking at FMODE_LSEEK in file->f_mode. diff --git a/drivers/gpu/drm/drm_file.c b/drivers/gpu/drm/drm_file.c index ed25168619fc..dc7d2e5b16c8 100644 --- a/drivers/gpu/drm/drm_file.c +++ b/drivers/gpu/drm/drm_file.c @@ -552,8 +552,7 @@ EXPORT_SYMBOL(drm_release_noglobal); * Since events are used by the KMS API for vblank and page flip completion this * means all modern display drivers must use it. * - * @offset is ignored, DRM events are read like a pipe. Therefore drivers also - * must set the &file_operation.llseek to no_llseek(). Polling support is + * @offset is ignored, DRM events are read like a pipe. Polling support is * provided by drm_poll(). * * This function will only ever read a full event. Therefore userspace must diff --git a/fs/file_table.c b/fs/file_table.c index 0658b822beeb..5727a63a7b67 100644 --- a/fs/file_table.c +++ b/fs/file_table.c @@ -235,7 +235,7 @@ static struct file *alloc_file(const struct path *path, int flags, file->f_mapping = path->dentry->d_inode->i_mapping; file->f_wb_err = filemap_sample_wb_err(file->f_mapping); file->f_sb_err = file_sample_sb_err(file); - if (fop->llseek && fop->llseek != no_llseek) + if (fop->llseek) file->f_mode |= FMODE_LSEEK; if ((file->f_mode & FMODE_READ) && likely(fop->read || fop->read_iter)) diff --git a/fs/open.c b/fs/open.c index 4488bd77c390..07c332753a36 100644 --- a/fs/open.c +++ b/fs/open.c @@ -860,8 +860,6 @@ static int do_dentry_open(struct file *f, f->f_mode |= FMODE_CAN_WRITE; if ((f->f_mode & FMODE_LSEEK) && !f->f_op->llseek) f->f_mode &= ~FMODE_LSEEK; - if ((f->f_mode & FMODE_LSEEK) && f->f_op->llseek == no_llseek) - f->f_mode &= ~FMODE_LSEEK; if (f->f_mapping->a_ops && f->f_mapping->a_ops->direct_IO) f->f_mode |= FMODE_CAN_ODIRECT; diff --git a/fs/read_write.c b/fs/read_write.c index d94b6dbba6f9..6b2849b34781 100644 --- a/fs/read_write.c +++ b/fs/read_write.c @@ -227,12 +227,6 @@ loff_t noop_llseek(struct file *file, loff_t offset, int whence) } EXPORT_SYMBOL(noop_llseek); -loff_t no_llseek(struct file *file, loff_t offset, int whence) -{ - return -ESPIPE; -} -EXPORT_SYMBOL(no_llseek); - loff_t default_llseek(struct file *file, loff_t offset, int whence) { struct inode *inode = file_inode(file); diff --git a/include/linux/fs.h b/include/linux/fs.h index 9ad5e3520fae..294932167335 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -3022,7 +3022,7 @@ extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out, extern void file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping); extern loff_t noop_llseek(struct file *file, loff_t offset, int whence); -extern loff_t no_llseek(struct file *file, loff_t offset, int whence); +#define no_llseek NULL extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize); extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence); extern loff_t generic_file_llseek_size(struct file *file, loff_t offset, diff --git a/kernel/bpf/bpf_iter.c b/kernel/bpf/bpf_iter.c index d5d96ceca105..8af0cbf9c0cd 100644 --- a/kernel/bpf/bpf_iter.c +++ b/kernel/bpf/bpf_iter.c @@ -81,10 +81,9 @@ static bool bpf_iter_support_resched(struct seq_file *seq) #define MAX_ITER_OBJECTS 1000000 /* bpf_seq_read, a customized and simpler version for bpf iterator. - * no_llseek is assumed for this file. * The following are differences from seq_read(): * . fixed buffer size (PAGE_SIZE) - * . assuming no_llseek + * . assuming NULL ->llseek() * . stop() may call bpf program, handling potential overflow there */ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, -- cgit v1.2.3