lkml.org 
[lkml]   [2003]   [Sep]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
Subject[BUG/MEMLEAK?] struct pci_bus, child busses & bridges
Whilst working on what was supposed to be a fairly trivial patch, I 
stumbled across what looks to be a relatively significant bug in the way
pci bridges are handled. I could easily be wrong, as the pci code is
far from anything I'd be willing to call an area of expertise. That
said, my description of the problem follows:

I am trying to add a file to the pci bus (ie: /sysfs/devices/pciXXXX:XX)
directories in sysfs. This led to the discovery that struct pci_bus
(inlude/linux/pci.h) does not have an *actual* struct dev inside it,
rather a pointer to a struct dev. I found this interesting, as most
device-type-thingies have an honest to goodness struct dev, allowing the
use of container_of(), etc. A quick perusal of the code offered no
reason why struct pci_bus couldn't be changed to have the actual struct
dev inside it, saving us kmalloc'ing and kfree'ing these, and generally
keeping track of them. I was wrong.

In pci_alloc_child_bus (drivers/pci/probe.c), the child bus is allocated
and it's struct dev * is set to point to the struct dev belonging to the
bridge that this bus is 'on', or 'behind'. pci_alloc_child_bus is
called in 3 places: pci_add_new_bus and twice in pci_scan_bridge. The
calls in pci_scan_bridge allocate a new struct pci_bus, but then seem to
throw the references away, *without* freeing them.

It appears that these pseudo-busses are allocated and used as a kind of
hack, to allow devices to be parented to pci bridge devices. The
pci_dev for the bridge is allocated and initialized, then a child bus is
created, and it's *dev is pointed to the struct device belonging to the
pci bridge. There are no refcounts used for this, and it doesn't appear
to be cleaned up in any way. If anyone can offer any insight into this
problem, or show me why I'm wrong, it would be greatly appreciated.

Thanks!

-Matt

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:48    [W:0.031 / U:0.952 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site