Skip to contentSkip to conversation

usb: xhci: fix event trb handling missed

PATCH New Ran Wang 3 participants 17 messages Sep 22, 2020 – Nov 13, 2020

Subsystems (2)
lore
Share & tools
lore

Ran Wang

author
#0 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that function mishandle event TRB in next round call, then system hang due to BUG() checking.

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

---
 drivers/usb/host/xhci-ring.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)
drivers/usb/host/xhci-ring.c +14 -3
index 092ed6e..d77e058 100644
@@ -578,10 +578,13 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
 	int ret;
 	u32 trb_fields[4];
 	u64 val_64 = virt_to_phys(buffer);
+	void *last_transfer_trb_addr;
+	int available_length;
 
 	debug("dev=%p, pipe=%lx, buffer=%p, length=%d\n",
 		udev, pipe, buffer, length);
 
+	available_length = length;
 	ep_index = usb_pipe_ep_index(pipe);
 	virt_dev = ctrl->devs[slot_id];
 
@@ -701,7 +704,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
 		trb_fields[2] = length_field;
 		trb_fields[3] = field | (TRB_NORMAL << TRB_TYPE_SHIFT);
 
-		queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
+		last_transfer_trb_addr = queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
 
 		--num_trbs;
 
@@ -714,6 +717,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
 
 	giveback_first_trb(udev, ep_index, start_cycle, start_trb);
 
+again:
 	event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
 	if (!event) {
 		debug("XHCI bulk transfer timed out, aborting...\n");
@@ -722,14 +726,21 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
 		udev->act_len = 0;
 		return -ETIMEDOUT;
 	}
-	field = le32_to_cpu(event->trans_event.flags);
 
+	if ((void *)event->trans_event.buffer != last_transfer_trb_addr) {
+		available_length -=
+			(int)EVENT_TRB_LEN(le32_to_cpu(event->trans_event.transfer_len));
+		xhci_acknowledge_event(ctrl);
+		goto again;
+	}
+
+	field = le32_to_cpu(event->trans_event.flags);
 	BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
 	BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
 	BUG_ON(*(void **)(uintptr_t)le64_to_cpu(event->trans_event.buffer) -
 		buffer > (size_t)length);
 
-	record_transfer_result(udev, event, length);
+	record_transfer_result(udev, event, available_length);
 	xhci_acknowledge_event(ctrl);
 	xhci_inval_cache((uintptr_t)buffer, length);
 
Signature · 1 line
2.7.4
#1 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Quoted 11 lines · Ran Wang

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that function mishandle event TRB in next round call, then system hang due to BUG() checking.

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

Bin, can you please take a look at this one ? Thanks

#2 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Quoted 11 lines · Marek Vasut
Quoted 9 lines · Ran Wang

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that function mishandle event TRB in next round call, then system hang due to BUG() checking.

Quoted 4 lines
Quoted 2 lines

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

Bin, can you please take a look at this one ? Thanks

Sorry I missed this one. Will take a look at this soon.

Regards,
Bin

author
#3 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin,

Quoted 18 lines · Bin Meng

To: Marek Vasut <[email protected]>
Cc: Ran Wang <[email protected]>; U-Boot Mailing List <[email protected]>
Subject: Re: [PATCH] usb: xhci: fix event trb handling missed

Quoted 12 lines · Marek Vasut
Quoted 10 lines · Ran Wang

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that function mishandle event TRB in next round call, then system hang due to
BUG() checking.

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

Bin, can you please take a look at this one ? Thanks

Sorry I missed this one. Will take a look at this soon.

Kindly remind that this patch is to resolve issue I ever sent to community for help:
https://lists.denx.de/pipermail/u-boot/2020-September/426981.html

Thanks & Regards,
Ran

#4 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

In reply to Ran Wang

Hi Ran,

Quoted 2 lines · Ran Wang

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will

typo: function

Quoted 5 lines

send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that

I am not sure if it's fair to say this is Layerscape specific behavior. Based on the xHCI spec, the spec indicates 1 event trb should be generated when IOC/ISP flag is set to 1 or an error occurs. I will see if I can reproduce your issue on an x86 board.

Quoted 47 lines

function mishandle event TRB in next round call, then system hang due to BUG() checking.

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

---
 drivers/usb/host/xhci-ring.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)
drivers/usb/host/xhci-ring.c +5 -1
index 092ed6e..d77e058 100644
@@ -578,10 +578,13 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
        int ret;
        u32 trb_fields[4];
        u64 val_64 = virt_to_phys(buffer);
+       void *last_transfer_trb_addr;
+       int available_length;
 
        debug("dev=%p, pipe=%lx, buffer=%p, length=%d\n",
                udev, pipe, buffer, length);
 
+       available_length = length;
        ep_index = usb_pipe_ep_index(pipe);
        virt_dev = ctrl->devs[slot_id];
 
@@ -701,7 +704,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
                trb_fields[2] = length_field;
                trb_fields[3] = field | (TRB_NORMAL << TRB_TYPE_SHIFT);
 
-               queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
+               last_transfer_trb_addr = queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
 
                --num_trbs;
 
@@ -714,6 +717,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
 
        giveback_first_trb(udev, ep_index, start_cycle, start_trb);
 
+again:
        event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
        if (!event) {
                debug("XHCI bulk transfer timed out, aborting...\n");

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe, · line 730

Quoted 6 lines
drivers/usb/host/xhci-ring.c +1 -1
@@ -722,14 +726,21 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
                udev->act_len = 0;
                return -ETIMEDOUT;
        }
-       field = le32_to_cpu(event->trans_event.flags);
 
+       if ((void *)event->trans_event.buffer != last_transfer_trb_addr) {

This should be:

if ((void *)le64_to_cpu(event->trans_event.buffer) != last_transfer_trb_addr)

Quoted 7 lines
drivers/usb/host/xhci-ring.c +7 -0
+               available_length -=
+                       (int)EVENT_TRB_LEN(le32_to_cpu(event->trans_event.transfer_len));
+               xhci_acknowledge_event(ctrl);
+               goto again;
+       }
+
+       field = le32_to_cpu(event->trans_event.flags);

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe, · line 746

Quoted 10 lines
drivers/usb/host/xhci-ring.c +1 -1
        BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
        BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
        BUG_ON(*(void **)(uintptr_t)le64_to_cpu(event->trans_event.buffer) -
                buffer > (size_t)length);
 
-       record_transfer_result(udev, event, length);
+       record_transfer_result(udev, event, available_length);
        xhci_acknowledge_event(ctrl);
        xhci_inval_cache((uintptr_t)buffer, length);
 

Regards,
Bin

#5 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Ran,

Quoted 12 lines · Bin Meng

Hi Ran,

Quoted 2 lines · Ran Wang

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will

typo: function

Quoted 3 lines

send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

Quoted 6 lines
Quoted 2 lines

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that

I am not sure if it's fair to say this is Layerscape specific behavior. Based on the xHCI spec, the spec indicates 1 event trb should be generated when IOC/ISP flag is set to 1 or an error occurs.

Ah, ISP flag is set if the pipe is from an IN endpoint. Currently we have:

        /* Only set interrupt on short packet for IN endpoints */
        if (usb_pipein(pipe))
            field |= TRB_ISP;

Can you verify that if removing the above codes, and without your changes in this patch, the original issue can be resolved on LS1028?

Quoted 2 lines

I will see if I can reproduce your issue on an x86 board.

Note this patch does not apply on top of u-boot/master. Please rebase.

Quoted 52 lines
Quoted 52 lines

function mishandle event TRB in next round call, then system hang due to BUG() checking.

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

---
 drivers/usb/host/xhci-ring.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)
drivers/usb/host/xhci-ring.c +5 -2
index 092ed6e..d77e058 100644
@@ -578,10 +578,13 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
        int ret;
        u32 trb_fields[4];
        u64 val_64 = virt_to_phys(buffer);
+       void *last_transfer_trb_addr;
+       int available_length;
 
        debug("dev=%p, pipe=%lx, buffer=%p, length=%d\n",
                udev, pipe, buffer, length);
 
+       available_length = length;
        ep_index = usb_pipe_ep_index(pipe);
        virt_dev = ctrl->devs[slot_id];
 
@@ -701,7 +704,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
                trb_fields[2] = length_field;
                trb_fields[3] = field | (TRB_NORMAL << TRB_TYPE_SHIFT);
 
-               queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
+               last_transfer_trb_addr = queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
 
                --num_trbs;
 
@@ -714,6 +717,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
 
        giveback_first_trb(udev, ep_index, start_cycle, start_trb);
 
+again:
        event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
        if (!event) {
                debug("XHCI bulk transfer timed out, aborting...\n");
@@ -722,14 +726,21 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
                udev->act_len = 0;
                return -ETIMEDOUT;
        }
-       field = le32_to_cpu(event->trans_event.flags);
 
Quoted 23 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe, · line 730

Quoted 1 line
drivers/usb/host/xhci-ring.c +1 -0
+       if ((void *)event->trans_event.buffer != last_transfer_trb_addr) {

This should be:

if ((void *)le64_to_cpu(event->trans_event.buffer) != last_transfer_trb_addr)

Quoted 17 lines
drivers/usb/host/xhci-ring.c +8 -1
+               available_length -=
+                       (int)EVENT_TRB_LEN(le32_to_cpu(event->trans_event.transfer_len));
+               xhci_acknowledge_event(ctrl);
+               goto again;
+       }
+
+       field = le32_to_cpu(event->trans_event.flags);
        BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
        BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
        BUG_ON(*(void **)(uintptr_t)le64_to_cpu(event->trans_event.buffer) -
                buffer > (size_t)length);
 
-       record_transfer_result(udev, event, length);
+       record_transfer_result(udev, event, available_length);
        xhci_acknowledge_event(ctrl);
        xhci_inval_cache((uintptr_t)buffer, length);
 

Regards,
Bin

author
#6 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin,

Quoted 13 lines · Bin Meng

Hi Ran,

Quoted 9 lines · Bin Meng

Hi Ran,

Quoted 3 lines · Ran Wang

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will

typo: function

Got it.

Quoted 11 lines
Quoted 10 lines
Quoted 6 lines

send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that

I am not sure if it's fair to say this is Layerscape specific behavior. Based on the xHCI spec, the spec indicates 1 event trb should be generated when IOC/ISP flag is set to 1 or an error occurs.

Quoted 8 lines

Ah, ISP flag is set if the pipe is from an IN endpoint. Currently we have:

        /* Only set interrupt on short packet for IN endpoints */
        if (usb_pipein(pipe))
            field |= TRB_ISP;

Can you verify that if removing the above codes, and without your changes in this patch, the original issue can be resolved on LS1028?

Bingo, removing above codes can resolve my issue, too

Quoted 4 lines
Quoted 2 lines

I will see if I can reproduce your issue on an x86 board.

Note this patch does not apply on top of u-boot/master. Please rebase.

Sure, I can rebase my patch, but I am nor sure my solution is still worth:
xHCI says "The detection of a USB Short Packet (i.e. the actual number of bytes received was less than the expected number of bytes defined by the Transfer TRB) during a transfer does not necessarily generate an Event. "

So does it mean above workaround you suggest would not violate spec, either (although current Linux kernel driver still set ISP for this case, but may have a more robust mechanism for event TRB handling)?

Quoted 46 lines
Quoted 12 lines
Quoted 12 lines

function mishandle event TRB in next round call, then system hang due to
BUG() checking.

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

---
 drivers/usb/host/xhci-ring.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)
Quoted 5 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev,

Quoted 5 lines
(quoted diff) +0 -0

b/drivers/usb/host/xhci-ring.c index 092ed6e..d77e058 100644

drivers/usb/host/xhci-ring.c +0 -0
@@ -578,10 +578,13 @@ int xhci_bulk_tx(struct usb_device *udev,

unsigned long pipe,

Quoted 6 lines
Quoted 6 lines
drivers/usb/host/xhci-ring.c +2 -0
        int ret;
        u32 trb_fields[4];
        u64 val_64 = virt_to_phys(buffer);
+       void *last_transfer_trb_addr;
+       int available_length;
 
Quoted 8 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned

Quoted 8 lines
drivers/usb/host/xhci-ring.c +1 -0
        debug("dev=%p, pipe=%lx, buffer=%p, length=%d\n",
                udev, pipe, buffer, length);
 
+       available_length = length;
        ep_index = usb_pipe_ep_index(pipe);
        virt_dev = ctrl->devs[slot_id];
 
@@ -701,7 +704,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned

long pipe,

Quoted 13 lines
Quoted 13 lines
drivers/usb/host/xhci-ring.c +2 -1
                trb_fields[2] = length_field;
                trb_fields[3] = field | (TRB_NORMAL <<
 TRB_TYPE_SHIFT);
 
-               queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
+               last_transfer_trb_addr = queue_trb(ctrl, ring,
+ (num_trbs > 1), trb_fields);
 
                --num_trbs;
@@ -714,6 +717,7 @@ int xhci_bulk_tx(struct usb_device *udev,
 unsigned long pipe,
 
Quoted 20 lines
Quoted 7 lines
Quoted 7 lines
drivers/usb/host/xhci-ring.c +1 -0
        giveback_first_trb(udev, ep_index, start_cycle, start_trb);
 
+again:
        event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
        if (!event) {

debug("XHCI bulk transfer timed out,
aborting...\n"); @@ -722,14 +726,21 @@ int xhci_bulk_tx(struct

usb_device *udev, unsigned long pipe,

Quoted 12 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,

Quoted 7 lines
                udev->act_len = 0;
                return -ETIMEDOUT;
        }
drivers/usb/host/xhci-ring.c +2 -1
-       field = le32_to_cpu(event->trans_event.flags);
 
+       if ((void *)event->trans_event.buffer !=
+ last_transfer_trb_addr) {

This should be:

if ((void *)le64_to_cpu(event->trans_event.buffer) != last_transfer_trb_addr)

Got it, will update this in next version (if you think my solution is still acceptable for this issue).

Thanks & regards,
Ran

Quoted 22 lines
Quoted 2 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe, · line 736

Quoted 2 lines
drivers/usb/host/xhci-ring.c +2 -0
+               available_length -=
+

(int)EVENT_TRB_LEN(le32_to_cpu(event->trans_event.transfer_len));

Quoted 8 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,

Quoted 8 lines
drivers/usb/host/xhci-ring.c +5 -0
+               xhci_acknowledge_event(ctrl);
+               goto again;
+       }
+
+       field = le32_to_cpu(event->trans_event.flags);
        BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
        BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
        BUG_ON(*(void

**)(uintptr_t)le64_to_cpu(event->trans_event.buffer) -

Quoted 7 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe, · line 745

Quoted 7 lines
                buffer > (size_t)length);
drivers/usb/host/xhci-ring.c +1 -1
-       record_transfer_result(udev, event, length);
+       record_transfer_result(udev, event, available_length);
        xhci_acknowledge_event(ctrl);
        xhci_inval_cache((uintptr_t)buffer, length);

Regards,
Bin

#7 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Ran,

Quoted 33 lines · Ran Wang

Hi Bin,

Quoted 13 lines · Bin Meng

Hi Ran,

Quoted 9 lines · Bin Meng

Hi Ran,

Quoted 3 lines · Ran Wang

In functiion xhci_bulk_tx(), when buffer cross 64KB boundary, it will

typo: function

Got it.

Quoted 13 lines
Quoted 10 lines
Quoted 6 lines

send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that

I am not sure if it's fair to say this is Layerscape specific behavior. Based on the xHCI spec, the spec indicates 1 event trb should be generated when IOC/ISP flag is set to 1 or an error occurs.

Ah, ISP flag is set if the pipe is from an IN endpoint. Currently we have:

Quoted 8 lines
Quoted 6 lines
        /* Only set interrupt on short packet for IN endpoints */
        if (usb_pipein(pipe))
            field |= TRB_ISP;

Can you verify that if removing the above codes, and without your changes in this patch, the original issue can be resolved on LS1028?

Bingo, removing above codes can resolve my issue, too

Thank you for your testing.

Quoted 11 lines
Quoted 5 lines
Quoted 3 lines

I will see if I can reproduce your issue on an x86 board.

Note this patch does not apply on top of u-boot/master. Please rebase.

Sure, I can rebase my patch, but I am nor sure my solution is still worth:
xHCI says "The detection of a USB Short Packet (i.e. the actual number of bytes received was less than the expected number of bytes defined by the Transfer TRB) during a transfer does not necessarily generate an Event. "

Yes, that's what the spec says. So in your case, can you add some logs to verify that there is a transfer event trb generated by the xHC and the completion code is 13 (short packet)? You can add some debug output in record_transfer_result().

Quoted 3 lines

So does it mean above workaround you suggest would not violate spec, either (although current Linux kernel driver still set ISP for this case, but may have a more robust mechanism for event TRB handling)?

I need to do more testing to see if this is a LS1028 behavior or generic xHCI behavior, ie: on an x86 board. Do you happen to know if there is any errata document for LS1028 that is related to this?

Quoted 66 lines
Quoted 66 lines
Quoted 14 lines
Quoted 14 lines

function mishandle event TRB in next round call, then system hang due to
BUG() checking.

This patch adds a loop to make sure the event TRB for last Transfer TRB has to be handled in time.

---
 drivers/usb/host/xhci-ring.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)
Quoted 5 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev,

Quoted 5 lines
(quoted diff) +0 -0

b/drivers/usb/host/xhci-ring.c index 092ed6e..d77e058 100644

drivers/usb/host/xhci-ring.c +0 -0
@@ -578,10 +578,13 @@ int xhci_bulk_tx(struct usb_device *udev,

unsigned long pipe,

Quoted 6 lines
Quoted 6 lines
drivers/usb/host/xhci-ring.c +2 -0
        int ret;
        u32 trb_fields[4];
        u64 val_64 = virt_to_phys(buffer);
+       void *last_transfer_trb_addr;
+       int available_length;
 
Quoted 8 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned

Quoted 8 lines
drivers/usb/host/xhci-ring.c +1 -0
        debug("dev=%p, pipe=%lx, buffer=%p, length=%d\n",
                udev, pipe, buffer, length);
 
+       available_length = length;
        ep_index = usb_pipe_ep_index(pipe);
        virt_dev = ctrl->devs[slot_id];
 
@@ -701,7 +704,7 @@ int xhci_bulk_tx(struct usb_device *udev, unsigned

long pipe,

Quoted 13 lines
Quoted 13 lines
drivers/usb/host/xhci-ring.c +2 -1
                trb_fields[2] = length_field;
                trb_fields[3] = field | (TRB_NORMAL <<
 TRB_TYPE_SHIFT);
 
-               queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
+               last_transfer_trb_addr = queue_trb(ctrl, ring,
+ (num_trbs > 1), trb_fields);
 
                --num_trbs;
@@ -714,6 +717,7 @@ int xhci_bulk_tx(struct usb_device *udev,
 unsigned long pipe,
 
Quoted 7 lines
Quoted 7 lines
drivers/usb/host/xhci-ring.c +1 -0
        giveback_first_trb(udev, ep_index, start_cycle, start_trb);
 
+again:
        event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
        if (!event) {

debug("XHCI bulk transfer timed out,
aborting...\n"); @@ -722,14 +726,21 @@ int xhci_bulk_tx(struct

usb_device *udev, unsigned long pipe,

Quoted 10 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,

Quoted 7 lines
                udev->act_len = 0;
                return -ETIMEDOUT;
        }
drivers/usb/host/xhci-ring.c +2 -1
-       field = le32_to_cpu(event->trans_event.flags);
 
+       if ((void *)event->trans_event.buffer !=
+ last_transfer_trb_addr) {

This should be:

Quoted 27 lines
Quoted 2 lines
Quoted 2 lines

if ((void *)le64_to_cpu(event->trans_event.buffer) != last_transfer_trb_addr)

Got it, will update this in next version (if you think my solution is still acceptable for this issue).

Thanks & regards,
Ran

Quoted 19 lines
Quoted 2 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe, · line 736

Quoted 2 lines
drivers/usb/host/xhci-ring.c +2 -0
+               available_length -=
+

(int)EVENT_TRB_LEN(le32_to_cpu(event->trans_event.transfer_len));

Quoted 8 lines

drivers/usb/host/xhci-ring.c int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,

Quoted 8 lines
drivers/usb/host/xhci-ring.c +5 -0
+               xhci_acknowledge_event(ctrl);
+               goto again;
+       }
+
+       field = le32_to_cpu(event->trans_event.flags);
        BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
        BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
        BUG_ON(*(void

**)(uintptr_t)le64_to_cpu(event->trans_event.buffer) -

Quoted 7 lines
Quoted 7 lines
                buffer > (size_t)length);
drivers/usb/host/xhci-ring.c +1 -1
-       record_transfer_result(udev, event, length);
+       record_transfer_result(udev, event, available_length);
        xhci_acknowledge_event(ctrl);
        xhci_inval_cache((uintptr_t)buffer, length);

Regards,
Bin

author
#8 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin,

Quoted 2 lines · Bin Meng

Hi Ran,

<snip>

Quoted 34 lines
Quoted 13 lines
Quoted 13 lines
Quoted 10 lines
Quoted 6 lines

send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that

I am not sure if it's fair to say this is Layerscape specific behavior. Based on the xHCI spec, the spec indicates 1 event trb should be generated when IOC/ISP flag is set to 1 or an error occurs.

Ah, ISP flag is set if the pipe is from an IN endpoint. Currently we have:

Quoted 8 lines
Quoted 6 lines
        /* Only set interrupt on short packet for IN endpoints */
        if (usb_pipein(pipe))
            field |= TRB_ISP;

Can you verify that if removing the above codes, and without your changes in this patch, the original issue can be resolved on LS1028?

Bingo, removing above codes can resolve my issue, too

Thank you for your testing.

Quoted 10 lines
Quoted 5 lines
Quoted 3 lines

I will see if I can reproduce your issue on an x86 board.

Note this patch does not apply on top of u-boot/master. Please rebase.

Sure, I can rebase my patch, but I am nor sure my solution is still worth:
xHCI says "The detection of a USB Short Packet (i.e. the actual number of bytes received was less than the expected number of bytes defined by the Transfer TRB) during a transfer does not necessarily generate an Event.

Quoted 7 lines

"

Quoted 1 line

Yes, that's what the spec says. So in your case, can you add some logs to verify that there is a transfer event trb generated by the xHC and the completion code is 13 (short packet)? You can add some debug output in record_transfer_result().

Sure, let me try this can get back to you later.

Quoted 7 lines
Quoted 3 lines

So does it mean above workaround you suggest would not violate spec, either (although current Linux kernel driver still set ISP for this case, but may have a more robust mechanism for event TRB handling)?

I need to do more testing to see if this is a LS1028 behavior or generic xHCI behavior, ie: on an x86 board. Do you happen to know if there is any errata document for LS1028 that is related to this?

I don?t see any errata for LS1028A for this case.

Actually this issue only randomly happens in U-Boot + USB dongle (I have tried 3 different brands, such as TP-LINK UE300, etc). And U-Boot + USB mass storage devicecannot reproduce this issue.

Under Linux, no such issue observed neither (with those dongles), all of them works fine. That's why I decide to improve event TRB handling logic here.

Below is the U-Boot test commands I used (repeatedly do tftp file transfer), issue would happen in less 5 minutes. for your reference:

setenv tt 'usb reset&&setenv ethact asix_eth&&tftp a0000000 tmp/1MB_test_file'
setenv ttt 'run tt&&run tt&&run tt&&run tt&&run tt&&run tt&&run tt&&run tt&&run tt&&run tt'
setenv tttt 'run ttt&&run ttt&&run ttt&&run ttt&&run ttt&&run ttt&&run ttt&&run ttt&&run ttt'
setenv u 'run tttt&&run tttt&&run tttt&&run tttt&&run tttt&&run tttt&&run tttt&&run tttt&&run tttt&&run tttt&&run tttt' run u

Regards,
Ran

author
#9 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin,

Quoted 44 lines · Ran Wang

Hi Bin,

Quoted 2 lines · Bin Meng

Hi Ran,

<snip>

Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
Quoted 10 lines
Quoted 6 lines

send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that

I am not sure if it's fair to say this is Layerscape specific behavior. Based on the xHCI spec, the spec indicates 1 event trb should be generated when IOC/ISP flag is set to 1 or an error occurs.

Ah, ISP flag is set if the pipe is from an IN endpoint. Currently we have:

        /* Only set interrupt on short packet for IN endpoints */
        if (usb_pipein(pipe))
            field |= TRB_ISP;
Quoted 17 lines
Quoted 4 lines
Quoted 2 lines

Can you verify that if removing the above codes, and without your changes in this patch, the original issue can be resolved on LS1028?

Bingo, removing above codes can resolve my issue, too

Thank you for your testing.

Quoted 10 lines
Quoted 5 lines
Quoted 3 lines

I will see if I can reproduce your issue on an x86 board.

Note this patch does not apply on top of u-boot/master. Please rebase.

Sure, I can rebase my patch, but I am nor sure my solution is still worth:
xHCI says "The detection of a USB Short Packet (i.e. the actual number of bytes received was less than the expected number of bytes defined by the Transfer TRB) during a transfer does not necessarily generate

an Event.

Quoted 3 lines

"

Quoted 1 line
Quoted 6 lines
Quoted 4 lines

Yes, that's what the spec says. So in your case, can you add some logs to verify that there is a transfer event trb generated by the xHC and the completion code is 13 (short packet)? You can add some debug output in record_transfer_result().

Sure, let me try this can get back to you later.

Confirm completion code is 13 when issue happen (please search ' complete_code:13'):

dev=00000000fbb4f3c0, pipe=c0010283, buffer=00000000fbb4fd80, length=2048----------------------------buffer would cross 64KB boundary, so we will send request in more than 1 TRB, this is the key issue trigger condition xhci_bulk_tx()#0.1.running_total:0x280
xhci_bulk_tx()#0.2.trb_buff_len:0x280
xhci_bulk_tx()#0.3.running_total:0x280
xhci_bulk_tx()#0.4.num_trbs:0x2--------------------------2 Transfer TRB
xhci_bulk_tx()#0.5.running_total:0x10280
xhci_bulk_tx()#0.start_trb:0x00000000fbb47140
----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47140----------Assemble 1st Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb4fd80
xhci_bulk_tx()#0.trb_buff_len:0x280
xhci_bulk_tx()#0.running_total:0x280
xhci_bulk_tx()#0.length:0x800
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000
----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47150----------Assemble 2nd Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb50000
xhci_bulk_tx()#0.trb_buff_len:0x580
xhci_bulk_tx()#0.running_total:0x800
xhci_bulk_tx()#0.length:0x800
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000
xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb47140----------------handle event TRB for 1st Transfer TRB
xhci_bulk_tx()#0.event->trans_event.transfer_len:0xd000180
xhci_bulk_tx()#0.event->trans_event.flags:0x1058001
xhci_bulk_tx()#0.event->len:384
xhci_bulk_tx()#0.event->complete_code:13
xhci_bulk_tx()#1.field:0x1058001
xhci_bulk_tx()#2.TRB_TO_EP_ENDEX(field):0x4
xhci_bulk_tx()#3.ep_index:0x4----------------------------------------------------------xhci_bulk_tx() return

dev=00000000fbb4f3c0, pipe=c0018203, buffer=00000000fbb2f9c0, length=110----------an other call comming, and for different EP (ep_index = 05) xhci_bulk_tx()#0.1.running_total:0x640
xhci_bulk_tx()#0.2.trb_buff_len:0x640
xhci_bulk_tx()#0.3.running_total:0x640
xhci_bulk_tx()#0.4.num_trbs:0x1
xhci_bulk_tx()#0.5.running_total:0x640
xhci_bulk_tx()#0.start_trb:0x00000000fbb47610
----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47610---------queue 1st transfer TRB
xhci_bulk_tx()#0.addr:0xfbb2f9c0
xhci_bulk_tx()#0.trb_buff_len:0x6e
xhci_bulk_tx()#0.running_total:0x6e
xhci_bulk_tx()#0.length:0x6e
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000
xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb47150------------hand event TRB, note that this buffer is for last Transfer TRB in last call (see above '2nd Transfer TRB') xhci_bulk_tx()#0.event->trans_event.transfer_len:0x1000000
xhci_bulk_tx()#0.event->trans_event.flags:0x1058000
xhci_bulk_tx()#0.event->len:0
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#1.field:0x1058000
xhci_bulk_tx()#2.TRB_TO_EP_ENDEX(field):0x4----------------ep_index is 4 rather than 5, then cause BUG() xhci_bulk_tx()#3.ep_index:0x5-------------------------------------
ep_index mis-match wait for event again!!!!!!!!!----------------------------------
BUG at drivers/usb/host/xhci-ring.c:759/xhci_bulk_tx()!
BUG!
### ERROR ### Please RESET the board ###

Regards,
Ran

#10 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Ran,

Quoted 48 lines · Ran Wang

Hi Bin,

Quoted 44 lines · Ran Wang

Hi Bin,

Quoted 2 lines · Bin Meng

Hi Ran,

<snip>

Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
Quoted 10 lines
Quoted 6 lines

send request in more than 1 Transfer TRB by chaining them, but then handle only 1 event TRB to mark request completed.

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that

I am not sure if it's fair to say this is Layerscape specific behavior. Based on the xHCI spec, the spec indicates 1 event trb should be generated when IOC/ISP flag is set to 1 or an error occurs.

Ah, ISP flag is set if the pipe is from an IN endpoint. Currently we have:

        /* Only set interrupt on short packet for IN endpoints */
        if (usb_pipein(pipe))
            field |= TRB_ISP;
Quoted 17 lines
Quoted 4 lines
Quoted 2 lines

Can you verify that if removing the above codes, and without your changes in this patch, the original issue can be resolved on LS1028?

Bingo, removing above codes can resolve my issue, too

Thank you for your testing.

Quoted 10 lines
Quoted 5 lines
Quoted 3 lines

I will see if I can reproduce your issue on an x86 board.

Note this patch does not apply on top of u-boot/master. Please rebase.

Sure, I can rebase my patch, but I am nor sure my solution is still worth:
xHCI says "The detection of a USB Short Packet (i.e. the actual number of bytes received was less than the expected number of bytes defined by the Transfer TRB) during a transfer does not necessarily generate

an Event.

Quoted 3 lines

"

Quoted 1 line
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

Yes, that's what the spec says. So in your case, can you add some logs to verify that there is a transfer event trb generated by the xHC and the completion code is 13 (short packet)? You can add some debug output in record_transfer_result().

Sure, let me try this can get back to you later.

Confirm completion code is 13 when issue happen (please search ' complete_code:13'):

Thank you very much for your testing.

Quoted 46 lines

dev=00000000fbb4f3c0, pipe=c0010283, buffer=00000000fbb4fd80, length=2048----------------------------buffer would cross 64KB boundary, so we will send request in more than 1 TRB, this is the key issue trigger condition xhci_bulk_tx()#0.1.running_total:0x280
xhci_bulk_tx()#0.2.trb_buff_len:0x280
xhci_bulk_tx()#0.3.running_total:0x280
xhci_bulk_tx()#0.4.num_trbs:0x2--------------------------2 Transfer TRB
xhci_bulk_tx()#0.5.running_total:0x10280
xhci_bulk_tx()#0.start_trb:0x00000000fbb47140
----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47140----------Assemble 1st Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb4fd80
xhci_bulk_tx()#0.trb_buff_len:0x280
xhci_bulk_tx()#0.running_total:0x280
xhci_bulk_tx()#0.length:0x800
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000
----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47150----------Assemble 2nd Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb50000
xhci_bulk_tx()#0.trb_buff_len:0x580
xhci_bulk_tx()#0.running_total:0x800
xhci_bulk_tx()#0.length:0x800
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000
xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb47140----------------handle event TRB for 1st Transfer TRB
xhci_bulk_tx()#0.event->trans_event.transfer_len:0xd000180
xhci_bulk_tx()#0.event->trans_event.flags:0x1058001
xhci_bulk_tx()#0.event->len:384
xhci_bulk_tx()#0.event->complete_code:13
xhci_bulk_tx()#1.field:0x1058001
xhci_bulk_tx()#2.TRB_TO_EP_ENDEX(field):0x4
xhci_bulk_tx()#3.ep_index:0x4----------------------------------------------------------xhci_bulk_tx() return

dev=00000000fbb4f3c0, pipe=c0018203, buffer=00000000fbb2f9c0, length=110----------an other call comming, and for different EP (ep_index = 05) xhci_bulk_tx()#0.1.running_total:0x640
xhci_bulk_tx()#0.2.trb_buff_len:0x640
xhci_bulk_tx()#0.3.running_total:0x640
xhci_bulk_tx()#0.4.num_trbs:0x1
xhci_bulk_tx()#0.5.running_total:0x640
xhci_bulk_tx()#0.start_trb:0x00000000fbb47610
----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47610---------queue 1st transfer TRB
xhci_bulk_tx()#0.addr:0xfbb2f9c0
xhci_bulk_tx()#0.trb_buff_len:0x6e
xhci_bulk_tx()#0.running_total:0x6e
xhci_bulk_tx()#0.length:0x6e
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000
xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb47150------------hand event TRB, note that this buffer is for last Transfer TRB in last call (see above '2nd Transfer TRB') xhci_bulk_tx()#0.event->trans_event.transfer_len:0x1000000
xhci_bulk_tx()#0.event->trans_event.flags:0x1058000
xhci_bulk_tx()#0.event->len:0

Quoted 9 lines

xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#1.field:0x1058000 xhci_bulk_tx()#2.TRB_TO_EP_ENDEX(field):0x4----------------ep_index is 4 rather than 5, then cause BUG() xhci_bulk_tx()#3.ep_index:0x5-------------------------------------
ep_index mis-match wait for event again!!!!!!!!!----------------------------------
BUG at drivers/usb/host/xhci-ring.c:759/xhci_bulk_tx()!
BUG!
### ERROR ### Please RESET the board ###

Based on your log, it matches what I suspected and we are getting close to the root cause.

Could you please try the following simple test case to see if it can trigger the issue with a USB disk?
After U-Boot boot on LS1028, with a USB flash disk attached, then type "usb start" to see if the USB disk can be recognized by U-Boot?

commit 60e68ed667362870c20b36ae26dacc1af903e81e

Author: Bin Meng <[email protected]>
Date:   Tue Nov 10 16:19:06 2020 +0800

WIP: usb: A simple test case to trigger the TRB 64K boundary issue
with mass storage device

This should not be applied as it aims to provide a simple test case to trigger the TRB 64K boundary issue as reported by Ran Wang @ NXP.

common/usb_storage.c +0 -0
index ff25441..c8aec2e 100644
@@ -710,7 +710,15 @@ static int usb_stor_BBB_transport(struct scsi_cmd

*srb, struct us_data *us)

        int dir_in;
        int actlen, data_actlen;
        unsigned int pipe, pipein, pipeout;
+#if 0
        ALLOC_CACHE_ALIGN_BUFFER(struct umass_bbb_csw, csw, 1);
+#else
+       struct umass_bbb_csw *csw_org, *csw;
+       csw_org = malloc(0x10000 + UMASS_BBB_CSW_SIZE);
+       csw = (struct umass_bbb_csw *)roundup((ulong)csw_org, 0x10000);
+       csw = (struct umass_bbb_csw *)((ulong)csw - 2);
+       printf("csw org %p, adjusted to %p\n", csw_org, csw);
+#endif
 #ifdef BBB_XPORT_TRACE
        unsigned char *ptr;
        int index;
@@ -824,6 +832,7 @@ again:
                return USB_STOR_TRANSPORT_FAILED;
        }
+       free(csw_org);
        return result;

}

Regards,
Bin

author
#11 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin,

Quoted 9 lines · Bin Meng

Hi Ran,

Quoted 5 lines · Ran Wang

Hi Bin,

Quoted 1 line · Ran Wang

Hi Bin,

<snip>

Quoted 43 lines
Quoted 1 line

Based on your log, it matches what I suspected and we are getting close to the root cause.

Could you please try the following simple test case to see if it can trigger the issue with a USB disk?
After U-Boot boot on LS1028, with a USB flash disk attached, then type "usb start" to see if the USB disk can be recognized by U-Boot?

commit 60e68ed667362870c20b36ae26dacc1af903e81e

Author: Bin Meng <[email protected]>
Date:   Tue Nov 10 16:19:06 2020 +0800

WIP: usb: A simple test case to trigger the TRB 64K boundary issue with
mass storage device

This should not be applied as it aims to provide a simple test case to trigger the TRB 64K boundary issue as reported by Ran Wang @ NXP.

common/usb_storage.c index renamed from common/usb_storage.c +0 -0

ff25441..c8aec2e 100644

common/usb_storage.c +7 -0
@@ -710,7 +710,15 @@ static int usb_stor_BBB_transport(struct scsi_cmd
 *srb, struct us_data *us)
        int dir_in;
        int actlen, data_actlen;
        unsigned int pipe, pipein, pipeout;
+#if 0
        ALLOC_CACHE_ALIGN_BUFFER(struct umass_bbb_csw, csw, 1);
+#else
+       struct umass_bbb_csw *csw_org, *csw;
+       csw_org = malloc(0x10000 + UMASS_BBB_CSW_SIZE);
+       csw = (struct umass_bbb_csw *)roundup((ulong)csw_org, 0x10000);
+       csw = (struct umass_bbb_csw *)((ulong)csw - 2);
+       printf("csw org %p, adjusted to %p\n", csw_org, csw); #endif
 #ifdef BBB_XPORT_TRACE
        unsigned char *ptr;
        int index;

common/usb_storage.c again: · line 838

Quoted 7 lines
common/usb_storage.c +1 -0
@@ -824,6 +832,7 @@ again:
                return USB_STOR_TRANSPORT_FAILED;
        }
 
+       free(csw_org);
        return result;
 }
 

Issue cannot be reproduced on mass storage device with above hacking:
=>
resetting USB...
Bus usb3 at 3100000: Register 200017f NbrPorts 2
Starting the controller
USB XHCI 1.00
Bus usb3 at 3110000: Register 200017f NbrPorts 2
Starting the controller
USB XHCI 1.00
scanning bus usb3 at 3100000 for devices... record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
2 USB Device(s) found
scanning bus usb3 at 3110000 for devices... 1 USB Device(s) found
scanning usb for storage devices... 1 Storage Device(s) found
=>

One more thing I'd like to tell is that I found this issue would not happened with 'usb start' (even with USB dongle).

Below is the log I tested with ' usb read 0x82000000 0 300000', issue could not be reproduced for at least 10 minutes
...
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
...

Regards,
Ran

#12 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Ran,

Quoted 62 lines · Ran Wang

Hi Bin,

Quoted 9 lines · Bin Meng

Hi Ran,

Quoted 5 lines · Ran Wang

Hi Bin,

Quoted 1 line · Ran Wang

Hi Bin,

<snip>

Quoted 46 lines
Quoted 1 line

Based on your log, it matches what I suspected and we are getting close to the root cause.

Could you please try the following simple test case to see if it can trigger the issue with a USB disk?
After U-Boot boot on LS1028, with a USB flash disk attached, then type "usb start" to see if the USB disk can be recognized by U-Boot?

commit 60e68ed667362870c20b36ae26dacc1af903e81e

Author: Bin Meng <[email protected]>
Date:   Tue Nov 10 16:19:06 2020 +0800

WIP: usb: A simple test case to trigger the TRB 64K boundary issue with
mass storage device

This should not be applied as it aims to provide a simple test case to trigger the TRB 64K boundary issue as reported by Ran Wang @ NXP.

common/usb_storage.c index renamed from common/usb_storage.c +0 -0

ff25441..c8aec2e 100644

common/usb_storage.c +7 -0
@@ -710,7 +710,15 @@ static int usb_stor_BBB_transport(struct scsi_cmd
 *srb, struct us_data *us)
        int dir_in;
        int actlen, data_actlen;
        unsigned int pipe, pipein, pipeout;
+#if 0
        ALLOC_CACHE_ALIGN_BUFFER(struct umass_bbb_csw, csw, 1);
+#else
+       struct umass_bbb_csw *csw_org, *csw;
+       csw_org = malloc(0x10000 + UMASS_BBB_CSW_SIZE);
+       csw = (struct umass_bbb_csw *)roundup((ulong)csw_org, 0x10000);
+       csw = (struct umass_bbb_csw *)((ulong)csw - 2);
+       printf("csw org %p, adjusted to %p\n", csw_org, csw); #endif
 #ifdef BBB_XPORT_TRACE
        unsigned char *ptr;
        int index;
@@ -824,6 +832,7 @@ again:
                return USB_STOR_TRANSPORT_FAILED;
        }
 
Quoted 6 lines

common/usb_storage.c again: · line 838

Quoted 4 lines
common/usb_storage.c +1 -0
+       free(csw_org);
        return result;
 }
 

Issue cannot be reproduced on mass storage device with above hacking:

Thank you for the testing.

Quoted 89 lines

=>
resetting USB...
Bus usb3 at 3100000: Register 200017f NbrPorts 2
Starting the controller
USB XHCI 1.00
Bus usb3 at 3110000: Register 200017f NbrPorts 2
Starting the controller
USB XHCI 1.00
scanning bus usb3 at 3100000 for devices... record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
2 USB Device(s) found
scanning bus usb3 at 3110000 for devices... 1 USB Device(s) found
scanning usb for storage devices... 1 Storage Device(s) found
=>

One more thing I'd like to tell is that I found this issue would not happened with 'usb start' (even with USB dongle).

Below is the log I tested with ' usb read 0x82000000 0 300000', issue could not be reproduced for at least 10 minutes
...
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1
record_transfer_result():comp_code: 1

Quoted 8 lines

csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1

In the earlier log you sent, you have the following:

dev=00000000fbb4f3c0, pipe=c0010283, buffer=00000000fbb4fd80, length=2048----------------------------buffer would cross 64KB
boundary, so we will send request in more than 1 TRB, this is the key issue trigger condition
xhci_bulk_tx()#0.1.running_total:0x280 xhci_bulk_tx()#0.2.trb_buff_len:0x280 xhci_bulk_tx()#0.3.running_total:0x280 xhci_bulk_tx()#0.4.num_trbs:0x2--------------------------2 Transfer TRB xhci_bulk_tx()#0.5.running_total:0x10280 xhci_bulk_tx()#0.start_trb:0x00000000fbb47140 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47140----------Assemble 1st Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb4fd80
xhci_bulk_tx()#0.trb_buff_len:0x280
xhci_bulk_tx()#0.running_total:0x280
xhci_bulk_tx()#0.length:0x800
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47150----------Assemble 2nd Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb50000
xhci_bulk_tx()#0.trb_buff_len:0x580
xhci_bulk_tx()#0.running_total:0x800
xhci_bulk_tx()#0.length:0x800
xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000

Could you please specify where this output is for?

"xhci_bulk_tx()#0.length:0x800"

Regards,
Bin

author
#13 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin,

Quoted 201 lines · Bin Meng

Hi Ran,

Quoted 8 lines · Ran Wang

Hi Bin,

Quoted 4 lines · Bin Meng

Hi Ran,

On Tue, Nov 10, 2020 at 4:31 PM Ran Wang <[email protected]>

Quoted 30 lines · wrote:
Quoted 5 lines
Quoted 5 lines

Hi Bin,

Quoted 1 line · Ran Wang

Hi Bin,

<snip>

Quoted 22 lines
Quoted 1 line

Based on your log, it matches what I suspected and we are getting close to the root cause.

Could you please try the following simple test case to see if it can trigger the issue with a USB disk?
After U-Boot boot on LS1028, with a USB flash disk attached, then type "usb start" to see if the USB disk can be recognized by U-Boot?

commit 60e68ed667362870c20b36ae26dacc1af903e81e

Author: Bin Meng <[email protected]>
Date:   Tue Nov 10 16:19:06 2020 +0800

WIP: usb: A simple test case to trigger the TRB 64K boundary
issue with mass storage device

This should not be applied as it aims to provide a simple test case to trigger the TRB 64K boundary issue as reported by Ran Wang @ NXP.

Quoted 15 lines
Quoted 5 lines
common/usb_storage.c index renamed from common/usb_storage.c +0 -0

ff25441..c8aec2e 100644

common/usb_storage.c static int usb_stor_BBB_transport(struct · line 719

Quoted 10 lines
common/usb_storage.c +5 -0
@@ -710,7 +710,15 @@ static int usb_stor_BBB_transport(struct
 scsi_cmd *srb, struct us_data *us)
        int dir_in;
        int actlen, data_actlen;
        unsigned int pipe, pipein, pipeout;
+#if 0
        ALLOC_CACHE_ALIGN_BUFFER(struct umass_bbb_csw, csw, 1);
+#else
+       struct umass_bbb_csw *csw_org, *csw;
+       csw_org = malloc(0x10000 + UMASS_BBB_CSW_SIZE);
+       csw = (struct umass_bbb_csw *)roundup((ulong)csw_org,

0x10000);

Quoted 9 lines
Quoted 9 lines
common/usb_storage.c +2 -0
+       csw = (struct umass_bbb_csw *)((ulong)csw - 2);
+       printf("csw org %p, adjusted to %p\n", csw_org, csw); #endif
 #ifdef BBB_XPORT_TRACE
        unsigned char *ptr;
        int index;
@@ -824,6 +832,7 @@ again:
                return USB_STOR_TRANSPORT_FAILED;
        }
 
Quoted 6 lines

common/usb_storage.c again: · line 838

Quoted 4 lines
common/usb_storage.c +1 -0
+       free(csw_org);
        return result;
 }
 

Issue cannot be reproduced on mass storage device with above hacking:

Thank you for the testing.

Quoted 30 lines

=>
resetting USB...
Bus usb3 at 3100000: Register 200017f NbrPorts 2 Starting the controller USB XHCI 1.00 Bus usb3 at 3110000: Register 200017f NbrPorts 2 Starting the controller USB XHCI 1.00 scanning bus usb3 at 3100000 for devices... record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 13 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb53010, adjusted to 00000000fbb5fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
2 USB Device(s) found
scanning bus usb3 at 3110000 for devices... 1 USB Device(s) found
scanning usb for storage devices... 1 Storage Device(s) found
=>

Quoted 1 line

One more thing I'd like to tell is that I found this issue would not happened

with 'usb start' (even with USB dongle).

Quoted 55 lines

Below is the log I tested with ' usb read 0x82000000 0 300000', issue could not be reproduced for at least 10 minutes ...
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1

Quoted 8 lines

csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1
csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1

In the earlier log you sent, you have the following:

dev=00000000fbb4f3c0, pipe=c0010283, buffer=00000000fbb4fd80, length=2048----------------------------buffer would cross 64KB boundary, so we will send request in more than 1 TRB, this is the key issue trigger condition xhci_bulk_tx()#0.1.running_total:0x280 xhci_bulk_tx()#0.2.trb_buff_len:0x280 xhci_bulk_tx()#0.3.running_total:0x280 xhci_bulk_tx()#0.4.num_trbs:0x2--------------------------2 Transfer TRB xhci_bulk_tx()#0.5.running_total:0x10280 xhci_bulk_tx()#0.start_trb:0x00000000fbb47140 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47140--------- -Assemble
1st Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb4fd80
xhci_bulk_tx()#0.trb_buff_len:0x280 xhci_bulk_tx()#0.running_total:0x280
xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47150--------- -Assemble
2nd Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb50000
xhci_bulk_tx()#0.trb_buff_len:0x580 xhci_bulk_tx()#0.running_total:0x800
xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000

Quoted 1 line

Could you please specify where this output is for?

All this prints is within function xhci_bulk_tx(),

Let me open those verbose output and check if that crossing 64KB case has been triggered or not. But I have to get back to you later (tomorrow).

Regards,
Ran

Quoted 4 lines

"xhci_bulk_tx()#0.length:0x800"

Regards,
Bin

author
#14 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin

Quoted 3 lines · Ran Wang

Hi Bin,

<snip>

Quoted 46 lines
Quoted 46 lines
Quoted 5 lines

record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90,

Quoted 8 lines

adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1

In the earlier log you sent, you have the following:

dev=00000000fbb4f3c0, pipe=c0010283, buffer=00000000fbb4fd80, length=2048----------------------------buffer would cross 64KB boundary, so we will send request in more than 1 TRB, this is the key issue trigger condition xhci_bulk_tx()#0.1.running_total:0x280 xhci_bulk_tx()#0.2.trb_buff_len:0x280 xhci_bulk_tx()#0.3.running_total:0x280 xhci_bulk_tx()#0.4.num_trbs:0x2--------------------------2 Transfer TRB
xhci_bulk_tx()#0.5.running_total:0x10280 xhci_bulk_tx()#0.start_trb:0x00000000fbb47140 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47140-
--------
-Assemble
1st Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb4fd80 xhci_bulk_tx()#0.trb_buff_len:0x280 xhci_bulk_tx()#0.running_total:0x280 xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47150-
--------
-Assemble
2nd Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb50000 xhci_bulk_tx()#0.trb_buff_len:0x580 xhci_bulk_tx()#0.running_total:0x800 xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000

Quoted 7 lines
Quoted 1 line

Could you please specify where this output is for?

All this prints is within function xhci_bulk_tx(),

Let me open those verbose output and check if that crossing 64KB case has been triggered or not.
But I have to get back to you later (tomorrow).

With your test commit and my verbose print code, issue cannot be reproduced in mass storage usage (although crossing 64KB boundary condition has met), so looks like the only different to ethernet device is that if xHC would receive short packet:

=> pri t
t=usb read 0x82000000 0 300000
=> run t

usb read: device 0 block # 0, count 3145728 ...
usb_stor_BBB_transport()csw org 00000000fbb4ac90, adjusted to 00000000fbb4fffe
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0010203, buffer=00000000fbb1a9c0, length=0x1f xhci_bulk_tx()#0.4.num_trbs:0x1
++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49570 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49570 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb49570
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x3
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb1ab00, length=0x200 xhci_bulk_tx()#0.4.num_trbs:0x1
xhci_bulk_tx()#set TRB_ISP------------------------------------ ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49180 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49180 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb49180
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x2
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb4fffe, length=0xd-----------------buffer cross 64KB boundary xhci_bulk_tx()#0.4.num_trbs:0x2
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49190----------------1st transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49190
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb491a0----------------2nd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb491a0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb491a0---------------------event TRB that pointing to 2nd transfer TRB xhci_bulk_tx()#0.event->complete_code:1-----------------------------------------------------------------not short packet xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!
...
usb_stor_BBB_transport()csw org 00000000fbb4aea0, adjusted to 00000000fbb4fffe
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0010203, buffer=00000000fbb1ad00, length=0x1f xhci_bulk_tx()#0.4.num_trbs:0x1
++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49790 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49790 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb49790
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x3
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000883d8000, length=0x1e000-----------------buffer cross 64KB boundary xhci_bulk_tx()#0.4.num_trbs:0x3
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492a0----------------1st transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492a0
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492b0----------------2nd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492b0
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492c0----------------3rd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492c0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb492c0---------------------event TRB that pointing to 3rd transfer TRB xhci_bulk_tx()#0.event->complete_code:1-----------------------------------------------------------------not short packet xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb4fffe, length=0xd xhci_bulk_tx()#0.4.num_trbs:0x2
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492d0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492d0
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492e0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492e0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb492e0
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!

Regards
Ran

#15 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Ran,

Quoted 60 lines · Ran Wang

Hi Bin

Quoted 3 lines · Ran Wang

Hi Bin,

<snip>

Quoted 50 lines
Quoted 46 lines
Quoted 5 lines

record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90,

Quoted 8 lines

adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1

In the earlier log you sent, you have the following:

dev=00000000fbb4f3c0, pipe=c0010283, buffer=00000000fbb4fd80, length=2048----------------------------buffer would cross 64KB boundary, so we will send request in more than 1 TRB, this is the key issue trigger condition xhci_bulk_tx()#0.1.running_total:0x280 xhci_bulk_tx()#0.2.trb_buff_len:0x280 xhci_bulk_tx()#0.3.running_total:0x280 xhci_bulk_tx()#0.4.num_trbs:0x2--------------------------2 Transfer TRB
xhci_bulk_tx()#0.5.running_total:0x10280 xhci_bulk_tx()#0.start_trb:0x00000000fbb47140 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47140-
--------
-Assemble
1st Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb4fd80 xhci_bulk_tx()#0.trb_buff_len:0x280 xhci_bulk_tx()#0.running_total:0x280 xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47150-
--------
-Assemble
2nd Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb50000 xhci_bulk_tx()#0.trb_buff_len:0x580 xhci_bulk_tx()#0.running_total:0x800 xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000

Quoted 1 line

Could you please specify where this output is for?

All this prints is within function xhci_bulk_tx(),

Quoted 8 lines
Quoted 3 lines

Let me open those verbose output and check if that crossing 64KB case has been triggered or not.
But I have to get back to you later (tomorrow).

With your test commit and my verbose print code, issue cannot be reproduced in mass storage usage (although crossing 64KB boundary condition has met), so looks like the only different to ethernet device is that if xHC would receive short packet:

Thank you.

Quoted 61 lines

=> pri t
t=usb read 0x82000000 0 300000
=> run t

usb read: device 0 block # 0, count 3145728 ...
usb_stor_BBB_transport()csw org 00000000fbb4ac90, adjusted to 00000000fbb4fffe
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0010203, buffer=00000000fbb1a9c0, length=0x1f xhci_bulk_tx()#0.4.num_trbs:0x1
++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49570 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49570 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb49570
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x3
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb1ab00, length=0x200 xhci_bulk_tx()#0.4.num_trbs:0x1
xhci_bulk_tx()#set TRB_ISP------------------------------------ ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49180 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49180 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb49180
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x2
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb4fffe, length=0xd-----------------buffer cross 64KB boundary xhci_bulk_tx()#0.4.num_trbs:0x2
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49190----------------1st transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49190
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb491a0----------------2nd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb491a0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb491a0---------------------event TRB that pointing to 2nd transfer TRB xhci_bulk_tx()#0.event->complete_code:1-----------------------------------------------------------------not short packet xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!
...
usb_stor_BBB_transport()csw org 00000000fbb4aea0, adjusted to 00000000fbb4fffe
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0010203, buffer=00000000fbb1ad00, length=0x1f xhci_bulk_tx()#0.4.num_trbs:0x1
++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb49790 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49790 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb49790
xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x3
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000883d8000, length=0x1e000-----------------buffer cross 64KB boundary xhci_bulk_tx()#0.4.num_trbs:0x3
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492a0----------------1st transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492a0
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492b0----------------2nd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492b0
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492c0----------------3rd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492c0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb492c0---------------------event TRB that pointing to 3rd transfer TRB xhci_bulk_tx()#0.event->complete_code:1-----------------------------------------------------------------not short packet xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!
xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb4fffe, length=0xd xhci_bulk_tx()#0.4.num_trbs:0x2
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492d0

Quoted 8 lines

xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492d0
xhci_bulk_tx()#set TRB_ISP---------------------------------------------------------------------------------------Set ISP ++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492e0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492e0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb492e0 xhci_bulk_tx()#0.event->complete_code:1
xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!

Please rebase this patch against u-boot/master, and reword the following commit message to indicate that it was because of the ISP flag set on an input pipe. In the 64K boundary case, when the first TRB ends up with a short packet condition it will trigger an short packet code transfer event per ISP flag, and current codes will only handle the first transfer event.

Quoted 4 lines

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that function mishandle event TRB in next round call, then system hang due to BUG() checking.

Regards,
Bin

author
#16 lore raw

PATCH 1/1 usb: xhci: fix event trb handling missed

Hi Bin,

Quoted 198 lines · Bin Meng

Hi Ran,

Quoted 62 lines · Ran Wang

Hi Bin

Quoted 3 lines · Ran Wang

Hi Bin,

<snip>

Quoted 52 lines
Quoted 48 lines
Quoted 5 lines

record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90,

Quoted 8 lines

adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 csw org 00000000fbb4ec90, adjusted to 00000000fbb4fffe record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1 record_transfer_result():comp_code: 1

In the earlier log you sent, you have the following:

dev=00000000fbb4f3c0, pipe=c0010283, buffer=00000000fbb4fd80, length=2048----------------------------buffer would cross 64KB boundary, so we will send request in more than 1 TRB, this is the key issue trigger condition xhci_bulk_tx()#0.1.running_total:0x280 xhci_bulk_tx()#0.2.trb_buff_len:0x280 xhci_bulk_tx()#0.3.running_total:0x280 xhci_bulk_tx()#0.4.num_trbs:0x2--------------------------2 Transfer TRB
xhci_bulk_tx()#0.5.running_total:0x10280 xhci_bulk_tx()#0.start_trb:0x00000000fbb47140 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47 140-
--------
-Assemble
1st Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb4fd80 xhci_bulk_tx()#0.trb_buff_len:0x280 xhci_bulk_tx()#0.running_total:0x280 xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000 ----------xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb47 150-
--------
-Assemble
2nd Transfer TRB
xhci_bulk_tx()#0.addr:0xfbb50000 xhci_bulk_tx()#0.trb_buff_len:0x580 xhci_bulk_tx()#0.running_total:0x800 xhci_bulk_tx()#0.length:0x800 xhci_bulk_tx()#0.TRB_MAX_BUFF_SIZE:0x10000

Quoted 1 line

Could you please specify where this output is for?

All this prints is within function xhci_bulk_tx(),

Quoted 7 lines
Quoted 3 lines

Let me open those verbose output and check if that crossing 64KB case has been triggered or not.
But I have to get back to you later (tomorrow).

With your test commit and my verbose print code, issue cannot be reproduced in mass storage usage (although crossing 64KB boundary condition has met), so looks like the only different to ethernet device is that if

xHC would receive short packet:

Quoted 1 line

Thank you.

Quoted 10 lines

=> pri t
t=usb read 0x82000000 0 300000
=> run t

usb read: device 0 block # 0, count 3145728 ... usb_stor_BBB_transport()csw org 00000000fbb4ac90, adjusted to 00000000fbb4fffe xhci_bulk_tx()dev=00000000fbb49950, pipe=c0010203, buffer=00000000fbb1a9c0, length=0x1f xhci_bulk_tx()#0.4.num_trbs:0x1

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb4957

Quoted 11 lines

++++++++++++0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49570 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb 49570
xhci_bulk_tx()#0.event->complete_code:1 xhci_bulk_tx()#3.ep_index:0x3 xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb1ab00, length=0x200 xhci_bulk_tx()#0.4.num_trbs:0x1
xhci_bulk_tx()#set TRB_ISP------------------------------------

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb4918

Quoted 5 lines

++++++++++++0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49180 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb 49180
xhci_bulk_tx()#0.event->complete_code:1

Quoted 9 lines

xhci_bulk_tx()#3.ep_index:0x2 xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb4fffe, length=0xd-----------------buffer cross 64KB boundary
xhci_bulk_tx()#0.4.num_trbs:0x2
xhci_bulk_tx()#set TRB_ISP--------------------------------------------------------------- ------------------------Set ISP

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb4919

Quoted 6 lines

++++++++++++0----------------1st transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49190 xhci_bulk_tx()#set TRB_ISP--------------------------------------------------------------- ------------------------Set ISP

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb491a

Quoted 5 lines

++++++++++++0----------------2nd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb491a0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb 491a0---------------------event TRB that pointing to 2nd transfer TRB xhci_bulk_tx()#0.event->complete_code:1-------------------------------

Quoted 9 lines

----------------------------------not short packet xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!
...
usb_stor_BBB_transport()csw org 00000000fbb4aea0, adjusted to 00000000fbb4fffe xhci_bulk_tx()dev=00000000fbb49950, pipe=c0010203, buffer=00000000fbb1ad00, length=0x1f xhci_bulk_tx()#0.4.num_trbs:0x1

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb4979

Quoted 5 lines

++++++++++++0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb49790 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb 49790
xhci_bulk_tx()#0.event->complete_code:1

Quoted 9 lines

xhci_bulk_tx()#3.ep_index:0x3 xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000883d8000, length=0x1e000-----------------buffer cross 64KB boundary
xhci_bulk_tx()#0.4.num_trbs:0x3
xhci_bulk_tx()#set TRB_ISP--------------------------------------------------------------- ------------------------Set ISP

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492a

Quoted 6 lines

++++++++++++0----------------1st transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492a0 xhci_bulk_tx()#set TRB_ISP--------------------------------------------------------------- ------------------------Set ISP

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492b

Quoted 6 lines

++++++++++++0----------------2nd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492b0 xhci_bulk_tx()#set TRB_ISP--------------------------------------------------------------- ------------------------Set ISP

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492c

Quoted 6 lines

++++++++++++0----------------3rd transfer TRB queued xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492c0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb 492c0---------------------event TRB that pointing to 3rd transfer TRB xhci_bulk_tx()#0.event->complete_code:1------------------------------- ----------------------------------not short packet

Quoted 9 lines

xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!! xhci_bulk_tx()dev=00000000fbb49950, pipe=c0008283, buffer=00000000fbb4fffe, length=0xd xhci_bulk_tx()#0.4.num_trbs:0x2
xhci_bulk_tx()#set TRB_ISP--------------------------------------------------------------- ------------------------Set ISP

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492d

Quoted 6 lines

++++++++++++0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492d0 xhci_bulk_tx()#set TRB_ISP--------------------------------------------------------------- ------------------------Set ISP

++++++++++++xhci_bulk_tx()#0.&ring->enqueue->generic:0x00000000fbb492e

Quoted 7 lines

++++++++++++0 xhci_bulk_tx()#0.last_transfer_trb_addr:0x00000000fbb492e0 --------------xhci_bulk_tx()#0.event->trans_event.buffer:0x00000000fbb 492e0
xhci_bulk_tx()#0.event->complete_code:1 xhci_bulk_tx()#3.ep_index:0x2
num_trbs_count > 1!!!!!!!!!

Quoted 5 lines

Please rebase this patch against u-boot/master, and reword the following commit message to indicate that it was because of the ISP flag set on an input pipe. In the 64K boundary case, when the first TRB ends up with a short packet condition it will trigger an short packet code transfer event per ISP flag, and current codes will only handle the first transfer event.

Got it, will work out v2, thanks for the review.

Regards,
Ran

Quoted 8 lines
Quoted 5 lines

However, on Layerscape platforms (LS1028A, LS1088A, etc), we observe xhci controller will generated more than 1 event TRB sometimes, this cause that function mishandle event TRB in next round call, then system hang due to
BUG() checking.

Regards,
Bin