summaryrefslogtreecommitdiffstats
path: root/drivers/usb/host/ehci.h
diff options
context:
space:
mode:
authorHans de Goede <hdegoede@redhat.com>2014-09-20 16:51:24 +0200
committerMarek Vasut <marex@denx.de>2014-10-06 14:50:42 +0200
commit415548d88446134549917aae026f53dbbee36fd2 (patch)
treed9d6cc62cd25aca896ea106992565a4e6d2d5459 /drivers/usb/host/ehci.h
parentea7b30c58973d59396c71d5880a652639d479a68 (diff)
downloadtalos-obmc-uboot-415548d88446134549917aae026f53dbbee36fd2.tar.gz
talos-obmc-uboot-415548d88446134549917aae026f53dbbee36fd2.zip
usb: ehci: poll_int_queue check real qtd, not the overlay
When we first start an int queue, the qh's overlay area is all zeros. This gets filled by the hc with the actual qtd values as soon as it advances the queue, but we may call poll_int_queue before then, in which case we would think the transfer has completed as the hc has not yet copied the qt_token to the overlay, so the active flag is not set. This fixes this by checking the actual qtd token, rather then the overlay. This also fixes a (theoretical) race where we see the completion in the overlay and free and re-use the qtd before the hc has completed writing back the overlay to the actual qtd. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Diffstat (limited to 'drivers/usb/host/ehci.h')
0 files changed, 0 insertions, 0 deletions
OpenPOWER on IntegriCloud