The latter is called only when both ino and dentry are about to
be freed, so cleaning ->d_fsdata and ->dentry is pointless.
Acked-by: Ian Kent <[email protected]>
Signed-off-by: Al Viro <[email protected]>
split init_ino into new_ino and clean_ino; the former is
what used to be init_ino(NULL, sbi), the latter is for cases
where we passed non-NULL ino. Lose unused arguments.
Acked-by: Ian Kent <[email protected]>
Signed-off-by: Al Viro <[email protected]>
It's used only to pass the length of symlink body to
autofs4_get_inode() in autofs4_dir_symlink(). We can
bloody well set inode->i_size in autofs4_dir_symlink()
directly and be done with that.
Acked-by: Ian Kent <[email protected]>
Signed-off-by: Al Viro <[email protected]>
In all cases we'd set inf->mode to know value just before
passing it to autofs4_get_inode(). That kills the need
to store it in autofs_info and pass it to autofs_init_ino()
Acked-by: Ian Kent <[email protected]>
Signed-off-by: Al Viro <[email protected]>
Kill it. Mind you, it's been an obfuscated call of autofs4_init_ino()
ever since 2.3.99pre6-4...
Acked-by: Ian Kent <[email protected]>
Signed-off-by: Al Viro <[email protected]>
gets rid of all ->free()/->u.symlink machinery in autofs; we simply
keep symlink bodies in inode->i_private and free them in ->evict_inode().
Acked-by: Ian Kent <[email protected]>
Signed-off-by: Al Viro <[email protected]>
do_add_mount() and mnt_clear_expiry() are not needed outside of
namespace.c anymore, now that namei has finish_automount() to
use.
Signed-off-by: Al Viro <[email protected]>
That gets rid of the kludge in finish_automount() - we need
to keep refcount on the vfsmount as-is until we evict it from
expiry list.
Signed-off-by: Al Viro <[email protected]>
Instead of splitting refcount between (per-cpu) mnt_count
and (SMP-only) mnt_longrefs, make all references contribute
to mnt_count again and keep track of how many are longterm
ones.
Accounting rules for longterm count:
* 1 for each fs_struct.root.mnt
* 1 for each fs_struct.pwd.mnt
* 1 for having non-NULL ->mnt_ns
* decrement to 0 happens only under vfsmount lock exclusive
That allows nice common case for mntput() - since we can't drop the
final reference until after mnt_longterm has reached 0 due to the rules
above, mntput() can grab vfsmount lock shared and check mnt_longterm.
If it turns out to be non-zero (which is the common case), we know
that this is not the final mntput() and can just blindly decrement
percpu mnt_count. Otherwise we grab vfsmount lock exclusive and
do usual decrement-and-check of percpu mnt_count.
For fs_struct.c we have mnt_make_longterm() and mnt_make_shortterm();
namespace.c uses the latter in places where we don't already hold
vfsmount lock exclusive and opencodes a few remaining spots where
we need to manipulate mnt_longterm.
Note that we mostly revert the code outside of fs/namespace.c back
to what we used to have; in particular, normal code doesn't need
to care about two kinds of references, etc. And we get to keep
the optimization Nick's variant had bought us...
Signed-off-by: Al Viro <[email protected]>
Expiry-related code calls umount_tree() several times with
the same list to collect vfsmounts to. Which is fine, except
that umount_tree() implicitly assumed that the list would
be empty on each call - it moves the victims over there and
then iterates through the list kicking them out. It's *almost*
idempotent, so everything nearly worked. However, mnt->ghosts
handling (and thus expirability checks) had been broken - that
part was not idempotent...
The fix is trivial - use local temporary list, splice it to
the the collector list when we are through.
Signed-off-by: Al Viro <[email protected]>
do_lookup() has a path leading from LOOKUP_RCU case to non-RCU
crossing of mountpoints, which breaks things badly. If we
hit need_revalidate: and do nothing in there, we need to come
back into LOOKUP_RCU half of things, not to done: in non-RCU
one.
Signed-off-by: Al Viro <[email protected]>
cgroup can't use simple_lookup(), since that'd override its desired ->d_op.
Tested-by: Li Zefan <[email protected]>
Signed-off-by: Al Viro <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
here we actually *want* ->d_op for root; setting it allows to get rid
of kludge in v9fs_kill_super() since now we have proper ->d_release()
for root and don't need to call it manually.
Signed-off-by: Al Viro <[email protected]>
Coda ->d_revalidate() actually checks for root, ->d_delete() is irrelevant.
So we can use the same d_op for all coda dentries
Signed-off-by: Al Viro <[email protected]>
don't bother with lock_super() in fat_fill_super() callers, while
we are at it - there won't be any concurrency anyway.
Signed-off-by: Al Viro <[email protected]>
we shouldn't bugger off to userland when there still are
pending signals; among other things it makes e.g. SIGSEGV
triggered by failure to build a sigframe to be delivered
_now_ and not when we hit the next syscall or interrupt.
Signed-off-by: Al Viro <[email protected]>
Acked-by: Greg Ungerer <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
If we leave sigreturn via ret_from_signal, we end up with syscall
trace only on entry, leading to very unhappy strace, among other
things. Note that this means different behaviours for signals
delivered while we were in pagefault and for ones delivered while
we were in interrupt...
Signed-off-by: Al Viro <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
a) we should hold modifying regs->format until we know we *will* be
doing stack expansion; otherwise attacker can modify sigframe to
have wrong ->sc_formatvec and install SIGSEGV handler.
b) we should *not* mix copying saved extra stuff from userland with
expanding the stack; once we'd done that manual memmove, we'd better
not return to C, so cleanup is very hard to do. The easiest way
is to copy it on stack first, making sure we won't overwrite on stack
expansion. Fortunately that's easy to do...
Signed-off-by: Al Viro <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
Same principle as with the previous patch - do not destroy the
state if sigframe setup fails. Incidentally, it's actually
_less_ work - we don't need to go through adjust_stack dance
on failure if we don't touch regs->stkadj until we know we'd
written sigframe out.
Signed-off-by: Al Viro <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
If we'd failed in setup_frame(), we've no place to store
the original sigmask. It's not an unrecoverable situation -
we raise SIGSEGV, but that SIGSEGV might be successfully
handled (e.g. on altstack). In that case we really don't
want sa_mask of original signal permanently slapped on
the set of blocked signals.
Standard solution: have setup_frame()/setup_rt_frame()
report failure and don't mess with the signal-related
state if that has happened...
Signed-off-by: Al Viro <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
Instead of checking the return value of do_signal() we can just do
the work (raise SIGTRAP and clear SR.T1) directly in handle_signal(),
when setting the sigframe up. Simplifies the assembler glue and is
closer to the way we do it on other targets.
Note that do_delayed_trace does *not* disappear; it's still needed
to deal with single-stepping through syscall, since 68040 doesn't
raise the trace exception at all if the trap exception is pending.
We hit it after returning from sys_...() if TIF_DELAYED_TRACE is
set; all that has changed is that we don't reuse it for "single-step
into the handler" codepath.
As the result, do_signal() doesn't need to return anything anymore.
Signed-off-by: Al Viro <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
... and had been such since the introduction of get_signal_to_deliver()
Signed-off-by: Al Viro <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
We never needed that (->regs[2] is overwritten on return from syscall paths
with return value of syscall, so storing it there early made no sense) and
with new restart logics since d27240bf7e61d2656de18e158ec910a902030847 it
has become really bad - we lose the original syscall number before the
place where we decide that we might need a syscall restart.
Note that for child we do need the assignment to regs[2] - it won't go
through the normal return from syscall path.
[Ralf: Issue found and reported by Lluís; initial investigations by me;
bug finally found and patch by Al; testing by me and Lluís.]
Signed-off-by: Al Viro <[email protected]>
Tested-by: Lluís Batlle i Rossell <[email protected]>
Signed-off-by: Ralf Baechle <[email protected]>
Normal syscall audit doesn't catch 5th argument of syscall. It also
doesn't catch the contents of userland structures pointed to be
syscall argument, so for both old and new mmap(2) ABI it doesn't
record the descriptor we are mapping. For old one it also misses
flags.
Signed-off-by: Al Viro <[email protected]>
take setting s_bdev/s_mtd/s_devops to callers of logfs_get_sb_device(),
don't bother passing them separately
Signed-off-by: Al Viro <[email protected]>
eventual replacement for ->get_sb() - does *not* get vfsmount,
return ERR_PTR(error) or root of subtree to be mounted.
Signed-off-by: Al Viro <[email protected]>
Al Viro