On Sat, 8 Dec 2007, Dr. Peer Griebel wrote:
> Dear USB developers
>
> some time ago (3/4 year) I already reported my problem with the EHCI
> driver. There was no response to it. I would like to tell you that my
> problem still persists. I just tried it today using the current Gentoo
> kernel 2.6.23-gentoo-r3.
>
> There don't seem to be many people with this problem and it is not very
> urgent for me to get it resolved. But I do think that this problem
> should be resolved. It is not very easy to find out that there are
> "lost IAA" messages since only the debug version of the driver shows the
> message. The average user doesn't know about the debug version. The
> everage user only sees that his web cam does not work.
>
> I'm a software developer myself. But I don't know very much about USB.
> I would be very glad if I can help you in any way to resolve my problem.
On Sat, 8 Dec 2007, Christian Hoffmann wrote:
> Hi,
> I am running into a similar crash with a dvb t receiver in end_async_unlink as
> reported some month ago on this ML.
>
> PID: 0 TASK: ffffffff804b7870 CPU: 0 COMMAND: "swapper"
> #0 [ffffffff80624b10] machine_kexec at ffffffff8021c00d
> #1 [ffffffff80624bc0] crash_kexec at ffffffff80255dbe
> #2 [ffffffff80624c80] do_page_fault at ffffffff803fde21
> #3 [ffffffff80624d90] error_exit at ffffffff803fc0fd
> [exception RIP: end_unlink_async+34]
> RIP: ffffffff8811d20a RSP: ffffffff80624e48 RFLAGS: 00010202
> RAX: 0000000000010029 RBX: ffff81007e098178 RCX: 0000000000000000
> RDX: 0000000000008028 RSI: ffffffff8811fe7e RDI: 0000000000000080
> RBP: 0000000000000000 R8: 0000000000000000 R9: ffffffff8047dc64
> R10: ffff8100022ffc48 R11: 0000000000000246 R12: 0000000000000000
> R13: ffff81007e098178 R14: ffff81007e098000 R15: 0000000000000000
> ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018
> #4 [ffffffff80624e60] ehci_work at ffffffff8811d78b
> #5 [ffffffff80624ed0] ehci_irq at ffffffff88120d0f
> #6 [ffffffff80624f20] usb_hcd_irq at ffffffff880ded82
> #7 [ffffffff80624f30] handle_IRQ_event at ffffffff802603e8
> #8 [ffffffff80624f60] handle_fasteoi_irq at ffffffff802618b6
> #9 [ffffffff80624f90] do_IRQ at ffffffff8020c136
> --- <IRQ stack> ---
> #10 [ffffffff805c7e88] ret_from_intr at ffffffff8020a141
> [exception RIP: acpi_safe_halt+40]
> RIP: ffffffff88003dc6 RSP: ffffffff805c7f30 RFLAGS: 00000246
> RAX: 0000000000000000 RBX: ffff81007e6760a8 RCX: ffffffff8047dde4
> RDX: 0000000000000007 RSI: 00000000000001c6 RDI: 0000000004000000
> RBP: 0000000000000000 R8: 0000000000000002 R9: ffffffff8047dc64
> R10: ffff8100022ffc48 R11: 0000000000000246 R12: ffffffff805c7f14
> R13: 0000000000001001 R14: ffffffff80332334 R15: 0000000000000000
> ORIG_RAX: ffffffffffffff86 CS: 0010 SS: 0018
> #11 [ffffffff805c7f30] acpi_processor_idle at ffffffff88003ff9
> #12 [ffffffff805c7f70] cpu_idle at ffffffff80208daf
> crash>
> I use 2.6.22.13-0.3-default on an Acer Laptop. ATI chipset. Anything
> else I can provide?
For both of you, here is a somewhat experimental patch you can try. It
probably won't fix the problem, but it might make the problem easier to
diagnose.
Along with this patch you should enable CONFIG_USB_DEBUG. If a problem
occurs, the contents of the files in /sys/class/usb_host/usb_hostN
(where N is the bus number of the EHCI controller in question) will be
of interest, especially the "async" and "registers" files.
This patch is for 2.6.24-rc4, but it ought to apply with little
difficulty to all recent kernels.
Alan Stern
Index: 2.6.24-rc4/drivers/usb/host/ehci-dbg.c
===================================================================
--- 2.6.24-rc4.orig/drivers/usb/host/ehci-dbg.c
+++ 2.6.24-rc4/drivers/usb/host/ehci-dbg.c
@@ -763,9 +763,7 @@ show_registers (struct class_device *cla
}
if (ehci->reclaim) {
- temp = scnprintf (next, size, "reclaim qh %p%s\n",
- ehci->reclaim,
- ehci->reclaim_ready ? " ready" : "");
+ temp = scnprintf(next, size, "reclaim qh %p\n", ehci->reclaim);
size -= temp;
next += temp;
}
Index: 2.6.24-rc4/drivers/usb/host/ehci-hcd.c
===================================================================
--- 2.6.24-rc4.orig/drivers/usb/host/ehci-hcd.c
+++ 2.6.24-rc4/drivers/usb/host/ehci-hcd.c
@@ -109,7 +109,7 @@ static const char hcd_name [] = "ehci_hc
#define EHCI_TUNE_MULT_TT 1
#define EHCI_TUNE_FLS 2 /* (small) 256 frame schedule */
-#define EHCI_IAA_JIFFIES (HZ/100) /* arbitrary; ~10 msec */
+#define EHCI_IAA_MSECS 10 /* arbitrary */
#define EHCI_IO_JIFFIES (HZ/10) /* io watchdog >
irq_thresh */
#define EHCI_ASYNC_JIFFIES (HZ/20) /* async idle timeout */
#define EHCI_SHRINK_JIFFIES (HZ/200) /* async qh unlink delay */
@@ -266,6 +266,7 @@ static void ehci_quiesce (struct ehci_hc
/*-------------------------------------------------------------------------*/
+static void end_unlink_async(struct ehci_hcd *ehci);
static void ehci_work(struct ehci_hcd *ehci);
#include "ehci-hub.c"
@@ -275,25 +276,41 @@ static void ehci_work(struct ehci_hcd *e
/*-------------------------------------------------------------------------*/
-static void ehci_watchdog (unsigned long param)
+static void ehci_iaa_watchdog(unsigned long param)
{
struct ehci_hcd *ehci = (struct ehci_hcd *) param;
unsigned long flags;
+ u32 status, cmd;
spin_lock_irqsave (&ehci->lock, flags);
+ WARN_ON(!ehci->reclaim);
- /* lost IAA irqs wedge things badly; seen with a vt8235 */
+ status = ehci_readl(ehci, &ehci->regs->status);
+ cmd = ehci_readl(ehci, &ehci->regs->command);
+ ehci_info(ehci, "IAA watchdog: status %x cmd %x\n", status, cmd);
+
+ /* lost IAA irqs wedge things badly; seen first with a vt8235 */
if (ehci->reclaim) {
- u32 status = ehci_readl(ehci, &ehci->regs->status);
if (status & STS_IAA) {
ehci_vdbg (ehci, "lost IAA\n");
COUNT (ehci->stats.lost_iaa);
ehci_writel(ehci, STS_IAA, &ehci->regs->status);
- ehci->reclaim_ready = 1;
}
+ ehci_writel(ehci, cmd & ~CMD_IAAD, &ehci->regs->command);
+ end_unlink_async(ehci);
}
- /* stop async processing after it's idled a bit */
+ spin_unlock_irqrestore(&ehci->lock, flags);
+}
+
+static void ehci_watchdog(unsigned long param)
+{
+ struct ehci_hcd *ehci = (struct ehci_hcd *) param;
+ unsigned long flags;
+
+ spin_lock_irqsave(&ehci->lock, flags);
+
+ /* stop async processing after it's idled a bit */
if (test_bit (TIMER_ASYNC_OFF, &ehci->actions))
start_unlink_async (ehci, ehci->async);
@@ -363,8 +380,6 @@ static void ehci_port_power (struct ehci
static void ehci_work (struct ehci_hcd *ehci)
{
timer_action_done (ehci, TIMER_IO_WATCHDOG);
- if (ehci->reclaim_ready)
- end_unlink_async (ehci);
/* another CPU may drop ehci->lock during a schedule scan while
* it reports urb completions. this flag guards against bogus
@@ -399,6 +414,7 @@ static void ehci_stop (struct usb_hcd *h
/* no more interrupts ... */
del_timer_sync (&ehci->watchdog);
+ del_timer_sync(&ehci->iaa_watchdog);
spin_lock_irq(&ehci->lock);
if (HC_IS_RUNNING (hcd->state))
@@ -447,6 +463,10 @@ static int ehci_init(struct usb_hcd *hcd
ehci->watchdog.function = ehci_watchdog;
ehci->watchdog.data = (unsigned long) ehci;
+ init_timer(&ehci->iaa_watchdog);
+ ehci->iaa_watchdog.function = ehci_iaa_watchdog;
+ ehci->iaa_watchdog.data = (unsigned long) ehci;
+
/*
* hw default: 1K periodic list heads, one per frame.
* periodic_size can shrink by USBCMD update if hcc_params allows.
@@ -463,7 +483,6 @@ static int ehci_init(struct usb_hcd *hcd
ehci->i_thresh = 2 + HCC_ISOC_THRES(hcc_params);
ehci->reclaim = NULL;
- ehci->reclaim_ready = 0;
ehci->next_uframe = -1;
/*
@@ -654,8 +673,7 @@ static irqreturn_t ehci_irq (struct usb_
/* complete the unlinking of some qh [4.15.2.3] */
if (status & STS_IAA) {
COUNT (ehci->stats.reclaim);
- ehci->reclaim_ready = 1;
- bh = 1;
+ end_unlink_async(ehci);
}
/* remote wakeup [4.3.1] */
@@ -761,10 +779,16 @@ static int ehci_urb_enqueue (
static void unlink_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
{
- /* if we need to use IAA and it's busy, defer */
- if (qh->qh_state == QH_STATE_LINKED
- && ehci->reclaim
- && HC_IS_RUNNING (ehci_to_hcd(ehci)->state)) {
+ /* failfast */
+ if (!HC_IS_RUNNING(ehci_to_hcd(ehci)->state))
+ end_unlink_async(ehci);
+
+ /* if it's not linked then there's nothing to do */
+ if (qh->qh_state != QH_STATE_LINKED)
+ ;
+
+ /* defer till later if busy */
+ else if (ehci->reclaim) {
struct ehci_qh *last;
for (last = ehci->reclaim;
@@ -774,12 +798,8 @@ static void unlink_async (struct ehci_hc
qh->qh_state = QH_STATE_UNLINK_WAIT;
last->reclaim = qh;
- /* bypass IAA if the hc can't care */
- } else if (!HC_IS_RUNNING (ehci_to_hcd(ehci)->state) && ehci->reclaim)
- end_unlink_async (ehci);
-
- /* something else might have unlinked the qh by now */
- if (qh->qh_state == QH_STATE_LINKED)
+ /* start IAA cycle */
+ } else
start_unlink_async (ehci, qh);
}
@@ -806,7 +826,19 @@ static int ehci_urb_dequeue(struct usb_h
qh = (struct ehci_qh *) urb->hcpriv;
if (!qh)
break;
- unlink_async (ehci, qh);
+ switch (qh->qh_state) {
+ case QH_STATE_LINKED:
+ case QH_STATE_COMPLETING:
+ unlink_async(ehci, qh);
+ break;
+ case QH_STATE_UNLINK:
+ case QH_STATE_UNLINK_WAIT:
+ /* already started */
+ break;
+ case QH_STATE_IDLE:
+ WARN_ON(1);
+ break;
+ }
break;
case PIPE_INTERRUPT:
@@ -898,6 +930,7 @@ rescan:
unlink_async (ehci, qh);
/* FALL THROUGH */
case QH_STATE_UNLINK: /* wait for hw to finish? */
+ case QH_STATE_UNLINK_WAIT:
idle_timeout:
spin_unlock_irqrestore (&ehci->lock, flags);
schedule_timeout_uninterruptible(1);
Index: 2.6.24-rc4/drivers/usb/host/ehci-hub.c
===================================================================
--- 2.6.24-rc4.orig/drivers/usb/host/ehci-hub.c
+++ 2.6.24-rc4/drivers/usb/host/ehci-hub.c
@@ -134,7 +134,7 @@ static int ehci_bus_suspend (struct usb_
}
ehci->command = ehci_readl(ehci, &ehci->regs->command);
if (ehci->reclaim)
- ehci->reclaim_ready = 1;
+ end_unlink_async(ehci);
ehci_work(ehci);
/* Unlike other USB host controller types, EHCI doesn't have
Index: 2.6.24-rc4/drivers/usb/host/ehci-pci.c
===================================================================
--- 2.6.24-rc4.orig/drivers/usb/host/ehci-pci.c
+++ 2.6.24-rc4/drivers/usb/host/ehci-pci.c
@@ -305,7 +305,7 @@ static int ehci_pci_resume(struct usb_hc
/* emptying the schedule aborts any urbs */
spin_lock_irq(&ehci->lock);
if (ehci->reclaim)
- ehci->reclaim_ready = 1;
+ end_unlink_async(ehci);
ehci_work(ehci);
spin_unlock_irq(&ehci->lock);
Index: 2.6.24-rc4/drivers/usb/host/ehci-q.c
===================================================================
--- 2.6.24-rc4.orig/drivers/usb/host/ehci-q.c
+++ 2.6.24-rc4/drivers/usb/host/ehci-q.c
@@ -973,7 +973,7 @@ static void end_unlink_async (struct ehc
struct ehci_qh *qh = ehci->reclaim;
struct ehci_qh *next;
- timer_action_done (ehci, TIMER_IAA_WATCHDOG);
+ iaa_watchdog_done(ehci);
// qh->hw_next = cpu_to_hc32(qh->qh_dma);
qh->qh_state = QH_STATE_IDLE;
@@ -983,7 +983,6 @@ static void end_unlink_async (struct ehc
/* other unlink(s) may be pending (in QH_STATE_UNLINK_WAIT) */
next = qh->reclaim;
ehci->reclaim = next;
- ehci->reclaim_ready = 0;
qh->reclaim = NULL;
qh_completions (ehci, qh);
@@ -1059,11 +1058,10 @@ static void start_unlink_async (struct e
return;
}
- ehci->reclaim_ready = 0;
cmd |= CMD_IAAD;
ehci_writel(ehci, cmd, &ehci->regs->command);
(void)ehci_readl(ehci, &ehci->regs->command);
- timer_action (ehci, TIMER_IAA_WATCHDOG);
+ iaa_watchdog_start(ehci);
}
/*-------------------------------------------------------------------------*/
Index: 2.6.24-rc4/drivers/usb/host/ehci.h
===================================================================
--- 2.6.24-rc4.orig/drivers/usb/host/ehci.h
+++ 2.6.24-rc4/drivers/usb/host/ehci.h
@@ -74,7 +74,6 @@ struct ehci_hcd { /* one per controlle
/* async schedule support */
struct ehci_qh *async;
struct ehci_qh *reclaim;
- unsigned reclaim_ready : 1;
unsigned scanning : 1;
/* periodic schedule support */
@@ -105,6 +104,7 @@ struct ehci_hcd { /* one per controlle
struct dma_pool *itd_pool; /* itd per iso urb */
struct dma_pool *sitd_pool; /* sitd per split iso urb */
+ struct timer_list iaa_watchdog;
struct timer_list watchdog;
unsigned long actions;
unsigned stamp;
@@ -140,9 +140,21 @@ static inline struct usb_hcd *ehci_to_hc
}
+static inline void
+iaa_watchdog_start(struct ehci_hcd *ehci)
+{
+ WARN_ON(timer_pending(&ehci->iaa_watchdog));
+ mod_timer(&ehci->iaa_watchdog,
+ jiffies + msecs_to_jiffies(EHCI_IAA_MSECS));
+}
+
+static inline void iaa_watchdog_done(struct ehci_hcd *ehci)
+{
+ del_timer(&ehci->iaa_watchdog);
+}
+
enum ehci_timer_action {
TIMER_IO_WATCHDOG,
- TIMER_IAA_WATCHDOG,
TIMER_ASYNC_SHRINK,
TIMER_ASYNC_OFF,
};
@@ -160,9 +172,6 @@ timer_action (struct ehci_hcd *ehci, enu
unsigned long t;
switch (action) {
- case TIMER_IAA_WATCHDOG:
- t = EHCI_IAA_JIFFIES;
- break;
case TIMER_IO_WATCHDOG:
t = EHCI_IO_JIFFIES;
break;
@@ -179,8 +188,7 @@ timer_action (struct ehci_hcd *ehci, enu
// 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 (action != TIMER_IAA_WATCHDOG
- && t > ehci->watchdog.expires
+ if (time_before_eq(t, ehci->watchdog.expires)
&& timer_pending (&ehci->watchdog))
return;
mod_timer (&ehci->watchdog, t);
-
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
