diff options
| author | Greg Kroah-Hartman <gregkh@suse.de> | 2008-06-18 16:39:50 -0700 |
|---|---|---|
| committer | Greg Kroah-Hartman <gregkh@suse.de> | 2008-06-18 16:39:50 -0700 |
| commit | ba9912c485a1886beaddd08feaadcb029e61de79 (patch) | |
| tree | 57d9443cae1cda478ee073c911dd356c6858ff02 | |
| parent | 3030fb3fc96c572dc77ac9b929aa38fb0e3a65ef (diff) | |
| download | patches-ba9912c485a1886beaddd08feaadcb029e61de79.tar.gz | |
2.6.26-rc6-git5 update and fix an ehci patch
| -rw-r--r-- | usb/usb-ehci-hcd-unlink-speedups.patch | 70 | ||||
| -rw-r--r-- | version | 2 |
2 files changed, 40 insertions, 32 deletions
diff --git a/usb/usb-ehci-hcd-unlink-speedups.patch b/usb/usb-ehci-hcd-unlink-speedups.patch index cc9310854e284a..e6370c06a4c62b 100644 --- a/usb/usb-ehci-hcd-unlink-speedups.patch +++ b/usb/usb-ehci-hcd-unlink-speedups.patch @@ -5,58 +5,59 @@ Subject: USB: ehci-hcd unlink speedups To: Greg KH <greg@kroah.com> Cc: linux-usb@vger.kernel.org, Alan Stern <stern@rowland.harvard.edu>, Leonid <leonidv11@gmail.com> Message-ID: <200806032221.56047.david-b@pacbell.net> -Content-Disposition: inline +From: David Brownell <dbrownell@users.sourceforge.net> -From: Alan Stern <stern@rowland.harvard.edu> +This patch fixes some performance bugs observed with some workloads +when unlinking EHCI queue header (QH) descriptors from the async ring +(control/bulk schedule). -This patch fixes a performance issue observed with some workloads -when unlinking EHCI queue header (QH) descriptors from the async -ring (control/bulk schedule). The mechanism intended to let an -empty QH stay scheduled for (only) a brief period, in case it's -quickly reused, was not working as intended. Sometimes the unlink -proceeded too quickly (which can be a strong negative effect); -sometimes it was too slow (wasting DMA cycles, usually a minor -issue except for bus contention and power usage). +The mechanism intended to defer unlinking an empty QH (so there is no +penalty in common cases where it's quickly reused) was not working as +intended. Sometimes the unlink was scheduled: -The fix replaces a simple counter with a timestamp derived from -the controller's microframe value. + - too quickly ... which can be a *strong* negative effect, since + that QH becomes unavailable for immediate re-use; -Finally, a logical error left over from the IAA watchdog-timer -conversion is corrected. Now the driver will always either unlink -an idle queue header or set up a timer to unlink it later. The old -code would sometimes fail to do either. + - too slowly ... wasting DMA cycles, usually a minor issue except + for increased bus contention and power usage; -[ dbrownell@users.sourceforge.net: cleanup timestamping + shrink timer ] +Plus there was an extreme case of "too slowly": a logical error in the +IAA watchdog-timer conversion meant that sometimes the unlink never +got scheduled. + +The fix replaces a simple counter with a timestamp derived from the +controller's 8 KHz microframe counter, and adjusts the timer usage +for some issues associated with HZ being less than 8K. + +(Based on a patch originally by Alan Stern, and good troubleshooting +from Leonid.) -Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> +Cc: Alan Stern <stern@rowland.harvard.edu> +Cc: Leonid <leonidv11@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> --- drivers/usb/host/ehci-hcd.c | 2 +- - drivers/usb/host/ehci-q.c | 15 ++++++++------- - drivers/usb/host/ehci.h | 6 +----- - 3 files changed, 10 insertions(+), 13 deletions(-) + drivers/usb/host/ehci-q.c | 17 +++++++++-------- + drivers/usb/host/ehci.h | 5 ++++- + 3 files changed, 14 insertions(+), 10 deletions(-) --- a/drivers/usb/host/ehci.h +++ b/drivers/usb/host/ehci.h -@@ -189,14 +189,10 @@ timer_action (struct ehci_hcd *ehci, enu +@@ -189,7 +189,10 @@ timer_action (struct ehci_hcd *ehci, enu break; // case TIMER_ASYNC_SHRINK: default: - t = EHCI_SHRINK_JIFFIES; -+ t = DIV_ROUND_UP(EHCI_SHRINK_FRAMES * HZ, 1000); ++ /* add a jiffie since we synch against the ++ * 8 KHz uframe counter. ++ */ ++ t = DIV_ROUND_UP(EHCI_SHRINK_FRAMES * HZ, 1000) + 1; break; } t += jiffies; -- // all timings except IAA watchdog can be overridden. -- // async queue SHRINK often precedes IAA. while it's ready -- // to go OFF neither can matter, and afterwards the IO -- // watchdog stops unless there's still periodic traffic. - if (time_before_eq(t, ehci->watchdog.expires) - && timer_pending (&ehci->watchdog)) - return; --- a/drivers/usb/host/ehci-hcd.c +++ b/drivers/usb/host/ehci-hcd.c @@ -84,7 +84,7 @@ static const char hcd_name [] = "ehci_hc @@ -80,7 +81,14 @@ Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> timer_action_done (ehci, TIMER_ASYNC_SHRINK); rescan: qh = ehci->async->qh_next.qh; -@@ -1148,12 +1147,14 @@ rescan: +@@ -1142,18 +1141,20 @@ rescan: + } + } + +- /* unlink idle entries, reducing HC PCI usage as well ++ /* unlink idle entries, reducing DMA usage as well + * as HCD schedule-scanning costs. delay for any qh + * we just scanned, there's a not-unusual case that it * doesn't stay idle for long. * (plus, avoids some kind of re-activation race.) */ @@ -1 +1 @@ -2.6.26-rc6-git4 +2.6.26-rc6-git5 |