Skip to contentSkip to conversation

xhci fixes for usb-linus

PATCH In review Mathias Nyman 4 participants 8 messages Jan 15, 2021 – Jan 19, 2021

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover xhci fixes for usb-linus

Hi Greg

A couple fixes for usb-linus.
Fixes one "xHCI died" case found when stress testing URB cancel on several devices at once, and one timing issue on Tegra xhci

JC Kuo (1):
xhci: tegra: Delay for disabling LFPS detector

Mathias Nyman (1):
xhci: make sure TRB is fully written before giving it to the
controller

 drivers/usb/host/xhci-ring.c  | 2 ++
 drivers/usb/host/xhci-tegra.c | 7 +++++++
 2 files changed, 9 insertions(+)
Signature · 1 line
2.25.1
#1 lore raw

PATCH 1/2 xhci: make sure TRB is fully written before giving it to the controller

Once the command ring doorbell is rung the xHC controller will parse all command TRBs on the command ring that have the cycle bit set properly.

If the driver just started writing the next command TRB to the ring when hardware finished the previous TRB, then HW might fetch an incomplete TRB as long as its cycle bit set correctly.

A command TRB is 16 bytes (128 bits) long.
Driver writes the command TRB in four 32 bit chunks, with the chunk containing the cycle bit last. This does however not guarantee that chunks actually get written in that order.

This was detected in stress testing when canceling URBs with several connected USB devices.
Two consecutive "Set TR Dequeue pointer" commands got queued right after each other, and the second one was only partially written when the controller parsed it, causing the dequeue pointer to be set to bogus values. This was seen as error messages:

"Mismatch between completed Set TR Deq Ptr command & xHCI internal state"

Solution is to add a write memory barrier before writing the cycle bit.

---
 drivers/usb/host/xhci-ring.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/usb/host/xhci-ring.c +2 -0
index 5677b81c0915..cf0c93a90200 100644
@@ -2931,6 +2931,8 @@ static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
 	trb->field[0] = cpu_to_le32(field1);
 	trb->field[1] = cpu_to_le32(field2);
 	trb->field[2] = cpu_to_le32(field3);
+	/* make sure TRB is fully written before giving it to the controller */
+	wmb();
 	trb->field[3] = cpu_to_le32(field4);
 
 	trace_xhci_queue_trb(ring, trb);
Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH 1/2] xhci: make sure TRB is fully written before giving it to the controller

Quoted 35 lines · Mathias Nyman

Once the command ring doorbell is rung the xHC controller will parse all command TRBs on the command ring that have the cycle bit set properly.

If the driver just started writing the next command TRB to the ring when hardware finished the previous TRB, then HW might fetch an incomplete TRB as long as its cycle bit set correctly.

A command TRB is 16 bytes (128 bits) long.
Driver writes the command TRB in four 32 bit chunks, with the chunk containing the cycle bit last. This does however not guarantee that chunks actually get written in that order.

This was detected in stress testing when canceling URBs with several connected USB devices.
Two consecutive "Set TR Dequeue pointer" commands got queued right after each other, and the second one was only partially written when the controller parsed it, causing the dequeue pointer to be set to bogus values. This was seen as error messages:

"Mismatch between completed Set TR Deq Ptr command & xHCI internal state"

Solution is to add a write memory barrier before writing the cycle bit.

---
 drivers/usb/host/xhci-ring.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/usb/host/xhci-ring.c +0 -0
index 5677b81c0915..cf0c93a90200 100644

drivers/usb/host/xhci-ring.c static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring, · line 2935

Quoted 5 lines
drivers/usb/host/xhci-ring.c +2 -0
@@ -2931,6 +2931,8 @@ static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
 	trb->field[0] = cpu_to_le32(field1);
 	trb->field[1] = cpu_to_le32(field2);
 	trb->field[2] = cpu_to_le32(field3);
+	/* make sure TRB is fully written before giving it to the controller */
+	wmb();

Have you tried the lighter barrier, dma_wmb()? IIRC, it exists for these exact cases...

drivers/usb/host/xhci-ring.c static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring, · line 2938

Quoted 3 lines
drivers/usb/host/xhci-ring.c +0 -0
 	trb->field[3] = cpu_to_le32(field4);
 
 	trace_xhci_queue_trb(ring, trb);

MBR, Sergei

#3 lore raw

RE: [PATCH 1/2] xhci: make sure TRB is fully written before giving it to the controller

From: Sergei Shtylyov

Quoted 34 lines

Sent: 15 January 2021 16:40

Quoted 30 lines · Mathias Nyman

Once the command ring doorbell is rung the xHC controller will parse all command TRBs on the command ring that have the cycle bit set properly.

If the driver just started writing the next command TRB to the ring when hardware finished the previous TRB, then HW might fetch an incomplete TRB as long as its cycle bit set correctly.

A command TRB is 16 bytes (128 bits) long.
Driver writes the command TRB in four 32 bit chunks, with the chunk containing the cycle bit last. This does however not guarantee that chunks actually get written in that order.

This was detected in stress testing when canceling URBs with several connected USB devices.
Two consecutive "Set TR Dequeue pointer" commands got queued right after each other, and the second one was only partially written when the controller parsed it, causing the dequeue pointer to be set to bogus values. This was seen as error messages:

"Mismatch between completed Set TR Deq Ptr command & xHCI internal state"

Solution is to add a write memory barrier before writing the cycle bit.

---
 drivers/usb/host/xhci-ring.c | 2 ++
 1 file changed, 2 insertions(+)
Quoted 12 lines

drivers/usb/host/xhci-ring.c static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring, · line 2935

Quoted 10 lines
drivers/usb/host/xhci-ring.c +2 -0
index 5677b81c0915..cf0c93a90200 100644
@@ -2931,6 +2931,8 @@ static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
 	trb->field[0] = cpu_to_le32(field1);
 	trb->field[1] = cpu_to_le32(field2);
 	trb->field[2] = cpu_to_le32(field3);
+	/* make sure TRB is fully written before giving it to the controller */
+	wmb();

Have you tried the lighter barrier, dma_wmb()? IIRC, it exists for these exact cases...

Isn't dma_wmb() needed between the last memory write and the io_write to the doorbell? Here we need to ensure the two memory writes aren't re-ordered.
Apart from alpha isn't a barrier() likely to be enough for that.
It is worth checking that the failing compiles didn't have the writes reordered.

David

Quoted 6 lines

drivers/usb/host/xhci-ring.c static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring, · line 2938

Quoted 4 lines
drivers/usb/host/xhci-ring.c +0 -0
 	trb->field[3] = cpu_to_le32(field4);
 
 	trace_xhci_queue_trb(ring, trb);

MBR, Sergei

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

#4 lore raw

Re: [PATCH 1/2] xhci: make sure TRB is fully written before giving it to the controller

Quoted 48 lines · David Laight

From: Sergei Shtylyov

Quoted 34 lines

Sent: 15 January 2021 16:40

Quoted 30 lines · Mathias Nyman

Once the command ring doorbell is rung the xHC controller will parse all command TRBs on the command ring that have the cycle bit set properly.

If the driver just started writing the next command TRB to the ring when hardware finished the previous TRB, then HW might fetch an incomplete TRB as long as its cycle bit set correctly.

A command TRB is 16 bytes (128 bits) long.
Driver writes the command TRB in four 32 bit chunks, with the chunk containing the cycle bit last. This does however not guarantee that chunks actually get written in that order.

This was detected in stress testing when canceling URBs with several connected USB devices.
Two consecutive "Set TR Dequeue pointer" commands got queued right after each other, and the second one was only partially written when the controller parsed it, causing the dequeue pointer to be set to bogus values. This was seen as error messages:

"Mismatch between completed Set TR Deq Ptr command & xHCI internal state"

Solution is to add a write memory barrier before writing the cycle bit.

---
 drivers/usb/host/xhci-ring.c | 2 ++
 1 file changed, 2 insertions(+)
Quoted 12 lines

drivers/usb/host/xhci-ring.c static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring, · line 2935

Quoted 10 lines
drivers/usb/host/xhci-ring.c +2 -0
index 5677b81c0915..cf0c93a90200 100644
@@ -2931,6 +2931,8 @@ static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
 	trb->field[0] = cpu_to_le32(field1);
 	trb->field[1] = cpu_to_le32(field2);
 	trb->field[2] = cpu_to_le32(field3);
+	/* make sure TRB is fully written before giving it to the controller */
+	wmb();

Have you tried the lighter barrier, dma_wmb()? IIRC, it exists for these exact cases...

Quoted 1 line

Isn't dma_wmb() needed between the last memory write and the io_write to the doorbell?

No.

Quoted 1 line

Here we need to ensure the two memory writes aren't re-ordered.

No, we need all 3 ring memory writes to be ordered such that they all happen before the 4th
write. It's not wonder this bug hasn't been noticed before -- x86 has strong write ordering unlike ARM/etc.

Quoted 1 line

Apart from alpha isn't a barrier() likely to be enough for that.

Not sure -- we don't have any barriers before the equivalents of a doorbell write
in e.g. the Renesas Ehter driver.

Quoted 1 line

It is worth checking that the failing compiles didn't have the writes reordered.

The writes are reordered not because of the compiler -- the read/write reordering is a
CPU feature (on at least non-x86). :-)

Quoted 6 lines

David

Quoted 4 lines
Quoted 4 lines
drivers/usb/host/xhci-ring.c +0 -0
 	trb->field[3] = cpu_to_le32(field4);
 
 	trace_xhci_queue_trb(ring, trb);

MBR, Sergei

#5 lore raw

Re: [PATCH 1/2] xhci: make sure TRB is fully written before giving it to the controller

Quoted 36 lines · Sergei Shtylyov
Quoted 35 lines · David Laight

From: Sergei Shtylyov

Quoted 34 lines

Sent: 15 January 2021 16:40

Quoted 30 lines · Mathias Nyman

Once the command ring doorbell is rung the xHC controller will parse all command TRBs on the command ring that have the cycle bit set properly.

If the driver just started writing the next command TRB to the ring when hardware finished the previous TRB, then HW might fetch an incomplete TRB as long as its cycle bit set correctly.

A command TRB is 16 bytes (128 bits) long.
Driver writes the command TRB in four 32 bit chunks, with the chunk containing the cycle bit last. This does however not guarantee that chunks actually get written in that order.

This was detected in stress testing when canceling URBs with several connected USB devices.
Two consecutive "Set TR Dequeue pointer" commands got queued right after each other, and the second one was only partially written when the controller parsed it, causing the dequeue pointer to be set to bogus values. This was seen as error messages:

"Mismatch between completed Set TR Deq Ptr command & xHCI internal state"

Solution is to add a write memory barrier before writing the cycle bit.

---
 drivers/usb/host/xhci-ring.c | 2 ++
 1 file changed, 2 insertions(+)
Quoted 13 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
drivers/usb/host/xhci-ring.c +0 -0
index 5677b81c0915..cf0c93a90200 100644
Quoted 9 lines
Quoted 8 lines

drivers/usb/host/xhci-ring.c static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring, · line 2935

Quoted 6 lines
drivers/usb/host/xhci-ring.c +2 -0
@@ -2931,6 +2931,8 @@ static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
 	trb->field[0] = cpu_to_le32(field1);
 	trb->field[1] = cpu_to_le32(field2);
 	trb->field[2] = cpu_to_le32(field3);
+	/* make sure TRB is fully written before giving it to the controller */
+	wmb();

Have you tried the lighter barrier, dma_wmb()? IIRC, it exists for these exact cases...

True, good point, dma_wmb() should be enough here.
In fact most other wmb()s in xhci could be turned into dma_wmb().

Looks like Greg already picked this so maybe a later patch to usb-next that does this wmb() -> dma_wmb() optimization where possible.

Quoted 3 lines
Quoted 1 line

Isn't dma_wmb() needed between the last memory write and the io_write to the doorbell?

   No.

Transfer trbs already have a wmb in giveback_first_trb() So no need in that case.

For command trbs it's unlikely but not impossible.
The issue we are solving here is xHC controller parsing two commands after a doorbell ring. First one was the intended, properly written command. Second was a out-of order partially written command. driver didn't even ring the doorbell for the second command yet.

There are a couple operations between trb last memory write and command doorbell ring. a wmb() in that place would solve a case where memory write is so out of order and delayed that xHC controller reads and reacts to the doorbell ring, and reads the command ring before the memory write to the command ring is done. Unlikely but not impossible.

No such issues seen so far, but maybe a dma_wmb() in xhci_ring_cmd_db() wouldn't hurt.

-Mathias

#6 lore raw

Re: [PATCH 1/2] xhci: make sure TRB is fully written before giving it to the controller

On Mon, Jan 18, 2021 at 5:05 AM Mathias Nyman
<[email protected]> wrote:
<>

Quoted 2 lines

True, good point, dma_wmb() should be enough here.
In fact most other wmb()s in xhci could be turned into dma_wmb().

FWIW I've confirmed in my testing that dma_wmb() does indeed also solve the problem.

#7 lore raw

PATCH 2/2 xhci: tegra: Delay for disabling LFPS detector

In reply to Mathias Nyman

From: JC Kuo <[email protected]>

Occasionally, we are seeing some SuperSpeed devices resumes right after being directed to U3. This commits add 500us delay to ensure LFPS detector is disabled before sending ACK to firmware.

[   16.099363] tegra-xusb 70090000.usb: entering ELPG
[   16.104343] tegra-xusb 70090000.usb: 2-1 isn't suspended: 0x0c001203
[   16.114576] tegra-xusb 70090000.usb: not all ports suspended: -16
[   16.120789] tegra-xusb 70090000.usb: entering ELPG failed

The register write passes through a few flop stages of 32KHz clock domain. NVIDIA ASIC designer reviewed RTL and suggests 500us delay.

---
 drivers/usb/host/xhci-tegra.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/usb/host/xhci-tegra.c +7 -0
index 934be1686352..50bb91b6a4b8 100644
@@ -623,6 +623,13 @@ static void tegra_xusb_mbox_handle(struct tegra_xusb *tegra,
 								     enable);
 			if (err < 0)
 				break;
+
+			/*
+			 * wait 500us for LFPS detector to be disabled before
+			 * sending ACK
+			 */
+			if (!enable)
+				usleep_range(500, 1000);
 		}
 
 		if (err < 0) {
Signature · 1 line
2.25.1