The patch titled
Subject: proc: tweak comments about 2 stage open and everything
has been added to the -mm tree. Its filename is
proc-tweak-comments-about-2-stage-open-and-everything.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/proc-tweak-comments-about-2-stage-open-and-everything.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/proc-tweak-comments-about-2-stage-open-and-everything.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/SubmitChecklist when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Alexey Dobriyan <adobriyan@xxxxxxxxx>
Subject: proc: tweak comments about 2 stage open and everything
Some comments were obsoleted since 05c0ae21c034a6f ("try a saner locking
for pde_opener...").
Some new comments added.
Some confusing comments replaced with equally confusing ones.
Link: http://lkml.kernel.org/r/20161029160231.GD1246@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/proc/inode.c | 29 +++++++++++++++++++++--------
1 file changed, 21 insertions(+), 8 deletions(-)
diff -puN fs/proc/inode.c~proc-tweak-comments-about-2-stage-open-and-everything fs/proc/inode.c
--- a/fs/proc/inode.c~proc-tweak-comments-about-2-stage-open-and-everything
+++ a/fs/proc/inode.c
@@ -138,6 +138,16 @@ static void unuse_pde(struct proc_dir_en
/* pde is locked */
static void close_pdeo(struct proc_dir_entry *pde, struct pde_opener *pdeo)
{
+ /*
+ * close() (proc_reg_release()) can't delete an entry and proceed:
+ * ->release hook needs to be available at the right moment.
+ *
+ * rmmod (remove_proc_entry() et al) can't delete an entry and proceed:
+ * "struct file" needs to be available at the right moment.
+ *
+ * Therefore, first process to enter this function does ->release() and
+ * signals its completion to the other process which does nothing.
+ */
if (pdeo->closing) {
/* somebody else is doing that, just wait */
DECLARE_COMPLETION_ONSTACK(c);
@@ -152,6 +162,7 @@ static void close_pdeo(struct proc_dir_e
file = pdeo->file;
pde->proc_fops->release(file_inode(file), file);
spin_lock(&pde->pde_unload_lock);
+ /* After ->release. */
list_del(&pdeo->lh);
if (pdeo->c)
complete(pdeo->c);
@@ -167,6 +178,8 @@ void proc_entry_rundown(struct proc_dir_
if (atomic_add_return(BIAS, &de->in_use) != BIAS)
wait_for_completion(&c);
+ /* ->pde_openers list can't grow from now on. */
+
spin_lock(&de->pde_unload_lock);
while (!list_empty(&de->pde_openers)) {
struct pde_opener *pdeo;
@@ -312,14 +325,15 @@ static int proc_reg_open(struct inode *i
struct pde_opener *pdeo;
/*
- * What for, you ask? Well, we can have open, rmmod, remove_proc_entry
- * sequence. ->release won't be called because ->proc_fops will be
- * cleared. Depending on complexity of ->release, consequences vary.
+ * Ensure that
+ * 1) PDE's ->release hook will be called no matter what
+ * either normally by close()/->release, or forcefully by
+ * rmmod/remove_proc_entry.
+ *
+ * 2) rmmod isn't blocked by opening file in /proc and sitting on
+ * the descriptor (including "rmmod foo </proc/foo" scenario).
*
- * We can't wait for mercy when close will be done for real, it's
- * deadlockable: rmmod foo </proc/foo . So, we're going to do ->release
- * by hand in remove_proc_entry(). For this, save opener's credentials
- * for later.
+ * Save every "struct file" with custom ->release hook.
*/
pdeo = kmalloc(sizeof(struct pde_opener), GFP_KERNEL);
if (!pdeo)
@@ -340,7 +354,6 @@ static int proc_reg_open(struct inode *i
pdeo->file = file;
pdeo->closing = false;
pdeo->c = NULL;
- /* Strictly for "too late" ->release in proc_reg_release(). */
spin_lock(&pde->pde_unload_lock);
list_add(&pdeo->lh, &pde->pde_openers);
spin_unlock(&pde->pde_unload_lock);
_
Patches currently in -mm which might be from adobriyan@xxxxxxxxx are
scripts-bloat-o-meter-fix-sigpipe.patch
kbuild-simpler-generation-of-assembly-constants.patch
proc-make-struct-pid_entry-len-unsigned.patch
proc-make-struct-struct-map_files_info-len-unsigned-int.patch
proc-just-list_del-struct-pde_opener.patch
proc-fix-type-of-struct-pde_opener-closing-field.patch
proc-kmalloc-struct-pde_opener.patch
proc-tweak-comments-about-2-stage-open-and-everything.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
