Fix kernel oops in pci_release_slot() when dbg() is enabled.
The dbg() in pci_release_slot() cause a kernel oops (invalid memory
access) because it accesses slot after it is released. We need to
access slot before calling kobject_put().
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
---
drivers/pci/slot.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
Index: linux-2.6.25-rc6/drivers/pci/slot.c
===================================================================
--- linux-2.6.25-rc6.orig/drivers/pci/slot.c
+++ linux-2.6.25-rc6/drivers/pci/slot.c
@@ -191,13 +191,14 @@ EXPORT_SYMBOL_GPL(pci_create_slot);
int pci_destroy_slot(struct pci_slot *slot)
{
+ dbg("%s: decreased refcount to %d on %x:%d\n", __func__,
+ atomic_read(&slot->kobj.kref.refcount) - 1, slot->bus->number,
+ slot->number);
+
down_write(&pci_bus_sem);
kobject_put(&slot->kobj);
up_write(&pci_bus_sem);
- dbg("%s: decreased refcount to %d on %x:%d\n", __func__,
- atomic_read(&slot->kobj.kref.refcount), slot->bus->number,
- slot->number);
return 0;
}
EXPORT_SYMBOL_GPL(pci_destroy_slot);
--
| Arjan van de Ven | [patch] Add basic sanity checks to the syscall execution patch |
| Matthew Wilcox | Re: AIM7 40% regression with 2.6.26-rc1 |
| Bart Van Assche | Integration of SCST in the mainstream Linux kernel |
| Greg Kroah-Hartman | [PATCH 005/196] Chinese: add translation of SubmittingDrivers |
git: | |
| Andy Whitcroft | Re: VCS comparison table |
| David | User's mailing list? And multiple cherry pick |
| Scott Chacon | Git Community Book |
| Mark Levedahl | Re: [PATCH] Teach remote machinery about remotes.default config variable |
| Marco Peereboom | Re: Real men don't attack straw men |
| Richard Stallman | Real men don't attack straw men |
| GVG GVG | ssh_exchange_identification: Connection closed by remote host |
| Tony Abernethy | Re: What is our ultimate goal?? |
| Arjan van de Ven | Re: [GIT]: Networking |
| Jeff Garzik | Re: [bug?] tg3: Failed to load firmware "tigon/tg3_tso.bin" |
| Denys Fedoryshchenko | packetloss, on e1000e worse than r8169? |
| Radu Rendec | Endianness problem with u32 classifier hash masks |
