On Fri, 8 Apr 2011, Joerg Roedel wrote:
> Booting 2.6.38 on my test machine produces a lockdep warning
> from the usb_amd_find_chipset_info() function:
>
> WARNING: at /data/lemmy/linux.trees.git/kernel/lockdep.c:2465 lockdep_trace_alloc+0x95/0xc2()
> Hardware name: Snook
> Modules linked in:
> Pid: 959, comm: work_for_cpu Not tainted 2.6.39-rc2+ #22
> Call Trace:
> [<ffffffff8103c0d4>] warn_slowpath_common+0x80/0x98
> [<ffffffff812387e6>] ? T.492+0x24/0x26
> [<ffffffff8103c101>] warn_slowpath_null+0x15/0x17
> [<ffffffff81068667>] lockdep_trace_alloc+0x95/0xc2
> [<ffffffff810ed9ac>] slab_pre_alloc_hook+0x18/0x3b
> [<ffffffff810ef227>] kmem_cache_alloc_trace+0x25/0xba
> [<ffffffff812387e6>] T.492+0x24/0x26
> [<ffffffff81238816>] pci_get_subsys+0x2e/0x73
> [<ffffffff8123886c>] pci_get_device+0x11/0x13
> [<ffffffff814082a9>] usb_amd_find_chipset_info+0x3f/0x18a
> ...
>
> It turns out that this function calls pci_get_device under a spin_lock
> with irqs disabled, but the pci_get_device function is only allowed in
> preemptible context.
>
> This patch fixes the warning by making all data-structure
> modifications on temporal storage and commiting this back
> into the visible structure at the end. While at it, this
> patch also moves the pci_dev_put calls out of the spinlocks
> because this function might sleep too.
I see only a couple of small flaws...
> diff --git a/drivers/usb/host/pci-quirks.c b/drivers/usb/host/pci-quirks.c
> index 1d586d4..dfc639a 100644
> --- a/drivers/usb/host/pci-quirks.c
> +++ b/drivers/usb/host/pci-quirks.c
> @@ -84,65 +84,91 @@ int usb_amd_find_chipset_info(void)
> {
> u8 rev = 0;
> unsigned long flags;
> + struct amd_chipset_info info;
> + int ret;
>
> spin_lock_irqsave(&amd_lock, flags);
>
> - amd_chipset.probe_count++;
> /* probe only once */
> - if (amd_chipset.probe_count > 1) {
> + if (amd_chipset.probe_count > 0) {
You need to increment probe_count here.
> spin_unlock_irqrestore(&amd_lock, flags);
> return amd_chipset.probe_result;
> }
> + info = amd_chipset;
What's the point of this line? You're just going to write over all the
data in info anyway, so it doesn't matter what amd_chipset contains. A
memset would work just as well.
> @@ -284,8 +310,10 @@ EXPORT_SYMBOL_GPL(usb_amd_quirk_pll_enable);
>
> void usb_amd_dev_put(void)
> {
> + struct pci_dev *nb, *smbus;
> unsigned long flags;
>
> +
Why add an extra blank line?
> spin_lock_irqsave(&amd_lock, flags);
>
> amd_chipset.probe_count--;
> @@ -294,20 +322,23 @@ void usb_amd_dev_put(void)
> return;
> }
>
> - if (amd_chipset.nb_dev) {
> - pci_dev_put(amd_chipset.nb_dev);
> - amd_chipset.nb_dev = NULL;
> - }
> - if (amd_chipset.smbus_dev) {
> - pci_dev_put(amd_chipset.smbus_dev);
> - amd_chipset.smbus_dev = NULL;
> - }
> + /* save them to pci_dev_put outside of spinlock */
> + nb = amd_chipset.nb_dev;
> + smbus = amd_chipset.smbus_dev;
> +
> + amd_chipset.nb_dev = NULL;
> + amd_chipset.smbus_dev = NULL;
> amd_chipset.nb_type = 0;
> amd_chipset.sb_type = 0;
> amd_chipset.isoc_reqs = 0;
> amd_chipset.probe_result = 0;
You could use memset instead. However, in reality it shouldn't be
necessary to set any of these things to 0.
>
> spin_unlock_irqrestore(&amd_lock, flags);
> +
> + if (nb)
> + pci_dev_put(nb);
> + if (smbus)
> + pci_dev_put(amd_chipset.smbus_dev);
> }
> EXPORT_SYMBOL_GPL(usb_amd_dev_put);
Alan Stern
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
