Re: [bug, 2.6.26-rc4/rc5] sporadic bootup crashes in blk_lookup_devt()/prepare_namespace()

["Vegard Nossum" <vegard.nossum@gmail.com>]

On 6/9/08, Vegard Nossum <vegard.nossum@gmail.com> wrote:
>  It seems that this list (block_class.devices) is protected by
>  block_class_lock in block/genhd.c. This list is only ever modified by
>  device_add() and device_del() in drivers/base/core.c. Both of those
>  are (only) protected by dev->class->sem, however. Is there a locking
>  mismatch here? But none of the locking code here seems to be changed
>  in years...

I think this seems correct.

Everywhere else where we traverse the struct class->devices list, they
have down(&class->sem); first and up(&class->sem); afterwards.

Commit fd04897bb20be29d60f7e426a053545aebeaa61a even has this hunk:
@@ -177,8 +177,7 @@ struct class {
        struct list_head        devices;
        struct list_head        interfaces;
        struct kset             class_dirs;
-       struct semaphore        sem;    /* locks both the children and interface
-
+       struct semaphore        sem; /* locks children, devices, interfaces */
        struct class_attribute          * class_attrs;
        struct class_device_attribute   * class_dev_attrs;
        struct device_attribute         * dev_attrs;


So why doesn't block/genhd.c do this too? It seems to me that the
mutex locking here is simply a remnant of old code that happened to
not crash in most cases by chance.


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036