]> git.baikalelectronics.ru Git - kernel.git/commit
Boot up with usermodehelper disabled
authorLinus Torvalds <torvalds@linux-foundation.org>
Thu, 4 Aug 2011 08:03:29 +0000 (22:03 -1000)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 4 Aug 2011 08:03:29 +0000 (22:03 -1000)
commite41c8904ac91aa4f52fc9ee2326e8c64b00b8cd1
tree58a6903344b8d9b2252144356a132a05a8359876
parent52a245b382c4683a9d6c99532299b3b70e206a1b
Boot up with usermodehelper disabled

The core device layer sends tons of uevent notifications for each device
it finds, and if the kernel has been built with a non-empty
CONFIG_UEVENT_HELPER_PATH that will make us try to execute the usermode
helper binary for all these events very early in the boot.

Not only won't the root filesystem even be mounted at that point, we
literally won't have necessarily even initialized all the process
handling data structures at that point, which causes no end of silly
problems even when the usermode helper doesn't actually succeed in
executing.

So just use our existing infrastructure to disable the usermodehelpers
to make the kernel start out with them disabled.  We enable them when
we've at least initialized stuff a bit.

Problems related to an uninitialized

init_ipc_ns.ids[IPC_SHM_IDS].rw_mutex

reported by various people.

Reported-by: Manuel Lauss <manuel.lauss@googlemail.com>
Reported-by: Richard Weinberger <richard@nod.at>
Reported-by: Marc Zyngier <maz@misterjones.org>
Acked-by: Kay Sievers <kay.sievers@vrfy.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
init/main.c
kernel/kmod.c