Re: [PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[Felipe Balbi]

Hi,

John Youn  writes:
>> Felipe Balbi  writes:
>>> At least macOS seems to be sending
>>> ClearFeature(ENDPOINT_HALT) to endpoints which
>>> aren't Halted. This makes DWC3's CLEARSTALL command
>>> time out which causes several issues for the driver.
>>>
>>> Instead, let's just return 0 and bail out early.
>>>
>>> Cc: 
>>> Signed-off-by: Felipe Balbi 
>>> ---
>>>
>>> this falls into "has never worked before" category, so I'll be sending
>>> it together with other patches for v4.11 merge window. Still, it's a
>>> valid bug that's likely needed for stable trees.
>>>
>>>  drivers/usb/dwc3/gadget.c | 5 +
>>>  1 file changed, 5 insertions(+)
>>>
>>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>>> index 6faf484e5dfc..0a664d8eba3f 100644
>>> --- a/drivers/usb/dwc3/gadget.c
>>> +++ b/drivers/usb/dwc3/gadget.c
>>> @@ -1379,6 +1379,9 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, 
>>> int value, int protocol)
>>> unsigned transfer_in_flight;
>>> unsigned started;
>>>
>>> +   if (dep->flags & DWC3_EP_STALL)
>>> +   return 0;
>>> +
>>> if (dep->number > 1)
>>> trb = dwc3_ep_prev_trb(dep, dep->trb_enqueue);
>>> else
>>> @@ -1400,6 +1403,8 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, 
>>> int value, int protocol)
>>> else
>>> dep->flags |= DWC3_EP_STALL;
>>> } else {
>>> +   if (!(dep->flags & DWC3_EP_STALL))
>>> +   return 0;
>>>
>>> ret = dwc3_send_clear_stall_ep_cmd(dep);
>>> if (ret)
>>
>> Reviving this old thread here. While $subject allowed dwc3 to work when
>> attached to macOS Host, I fear that we might have more issues than not
>> in the future. The reason is that USB20 spec allows hosts to use
>> ClearFeature(ENDPOINT_HALT) as a "Reset Data Toggle/SeqN" hint.
>>
>> With this, we're basically blocking that possibility. Still, without
>> $subject, ClearStall commands were timing out. I'll try to do a local
>> revert here and check what happens in this case, but would you have any
>> idea why ClearStall would time out like that?
>>
>
> Hi Felipe,
>
> I think Thinh Nguyen e-mailed you about this before saying it caused a
> regression for us. But he has not had time to look into it further and
> follow-up yet.
>
> No idea about the timing out on Mac though. We can try to reproduce
> this when we have some time and take a look. Do you have a USB trace?

not anymore, but as soon I have some free time, I can revert the patch
locally and try to reproduce. I'll also implement a little libusb-based
test to issue that request hundreds of times in a row and see if I can
force the problem to happen ;-)

Thanks for responding so quickly :-)

-- 
balbi
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[John Youn]

On 04/04/2017 01:09 AM, Felipe Balbi wrote:
>
> Hi John,
>
> Felipe Balbi  writes:
>> At least macOS seems to be sending
>> ClearFeature(ENDPOINT_HALT) to endpoints which
>> aren't Halted. This makes DWC3's CLEARSTALL command
>> time out which causes several issues for the driver.
>>
>> Instead, let's just return 0 and bail out early.
>>
>> Cc: 
>> Signed-off-by: Felipe Balbi 
>> ---
>>
>> this falls into "has never worked before" category, so I'll be sending
>> it together with other patches for v4.11 merge window. Still, it's a
>> valid bug that's likely needed for stable trees.
>>
>>  drivers/usb/dwc3/gadget.c | 5 +
>>  1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>> index 6faf484e5dfc..0a664d8eba3f 100644
>> --- a/drivers/usb/dwc3/gadget.c
>> +++ b/drivers/usb/dwc3/gadget.c
>> @@ -1379,6 +1379,9 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, int 
>> value, int protocol)
>>  unsigned transfer_in_flight;
>>  unsigned started;
>>
>> +if (dep->flags & DWC3_EP_STALL)
>> +return 0;
>> +
>>  if (dep->number > 1)
>>  trb = dwc3_ep_prev_trb(dep, dep->trb_enqueue);
>>  else
>> @@ -1400,6 +1403,8 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, int 
>> value, int protocol)
>>  else
>>  dep->flags |= DWC3_EP_STALL;
>>  } else {
>> +if (!(dep->flags & DWC3_EP_STALL))
>> +return 0;
>>
>>  ret = dwc3_send_clear_stall_ep_cmd(dep);
>>  if (ret)
>
> Reviving this old thread here. While $subject allowed dwc3 to work when
> attached to macOS Host, I fear that we might have more issues than not
> in the future. The reason is that USB20 spec allows hosts to use
> ClearFeature(ENDPOINT_HALT) as a "Reset Data Toggle/SeqN" hint.
>
> With this, we're basically blocking that possibility. Still, without
> $subject, ClearStall commands were timing out. I'll try to do a local
> revert here and check what happens in this case, but would you have any
> idea why ClearStall would time out like that?
>

Hi Felipe,

I think Thinh Nguyen e-mailed you about this before saying it caused a
regression for us. But he has not had time to look into it further and
follow-up yet.

No idea about the timing out on Mac though. We can try to reproduce
this when we have some time and take a look. Do you have a USB trace?

Regards,
John
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[Felipe Balbi]

Hi John,

Felipe Balbi  writes:
> At least macOS seems to be sending
> ClearFeature(ENDPOINT_HALT) to endpoints which
> aren't Halted. This makes DWC3's CLEARSTALL command
> time out which causes several issues for the driver.
>
> Instead, let's just return 0 and bail out early.
>
> Cc: 
> Signed-off-by: Felipe Balbi 
> ---
>
> this falls into "has never worked before" category, so I'll be sending
> it together with other patches for v4.11 merge window. Still, it's a
> valid bug that's likely needed for stable trees.
>
>  drivers/usb/dwc3/gadget.c | 5 +
>  1 file changed, 5 insertions(+)
>
> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
> index 6faf484e5dfc..0a664d8eba3f 100644
> --- a/drivers/usb/dwc3/gadget.c
> +++ b/drivers/usb/dwc3/gadget.c
> @@ -1379,6 +1379,9 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, int 
> value, int protocol)
>   unsigned transfer_in_flight;
>   unsigned started;
>  
> + if (dep->flags & DWC3_EP_STALL)
> + return 0;
> +
>   if (dep->number > 1)
>   trb = dwc3_ep_prev_trb(dep, dep->trb_enqueue);
>   else
> @@ -1400,6 +1403,8 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, int 
> value, int protocol)
>   else
>   dep->flags |= DWC3_EP_STALL;
>   } else {
> + if (!(dep->flags & DWC3_EP_STALL))
> + return 0;
>  
>   ret = dwc3_send_clear_stall_ep_cmd(dep);
>   if (ret)

Reviving this old thread here. While $subject allowed dwc3 to work when
attached to macOS Host, I fear that we might have more issues than not
in the future. The reason is that USB20 spec allows hosts to use
ClearFeature(ENDPOINT_HALT) as a "Reset Data Toggle/SeqN" hint.

With this, we're basically blocking that possibility. Still, without
$subject, ClearStall commands were timing out. I'll try to do a local
revert here and check what happens in this case, but would you have any
idea why ClearStall would time out like that?

-- 
balbi


signature.asc
Description: PGP signature

RE: [PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[Felipe Balbi]

Hi,

Felipe Balbi  writes:
>>> I've been looking at this and based on sniffer and dwc3 tracepoints, it
>>> seems like dwc3 is behaving properly. The real issue seems to be that
>>> g_mass_storage isn't queueing a new request to IN endpoint.
>>>
>>> I'll continue debugging this and try to find a solution that doesn't
>>> involve reverting $subject.
>>
>> oh no, wait. ep2out misses XferInProgress:
>>
>> file-storage-1592  [000] d..1   152.809922: dwc3_ep_queue: ep2out: req 
>> 88003cd6ee40 length 0/512 zsI ==> -115
>> file-storage-1592  [000] d..1   152.809931: dwc3_prepare_trb: ep2out: 
>> 3/8 trb 88003a196050 buf 2d5e4000 size 512 ctrl 0819 
>> (HlcS:sC:normal)
>> file-storage-1592  [000] d..1   152.809942: dwc3_gadget_ep_cmd: ep2out: 
>> cmd 'Update Transfer' [262151] params    --> status: 
>> Successful
>> file-storage-1592  [000]    152.809951: usb_ep_queue: ep2out: length 
>> 0/512 sgs 0/0 stream 0 zsI status -115 --> 0
>>  irq/34-dwc3-1593  [001] d..1   152.810212: dwc3_event: event 
>> (c040): ep0out: Transfer Complete [Setup Phase]
>>  irq/34-dwc3-1593  [001] d..1   152.810218: dwc3_ctrl_req: bRequestType 
>> 02 bRequest 01 wValue  wIndex 0002 wLength 0
>>  irq/34-dwc3-1593  [001] d..1   152.810228: __dwc3_gadget_ep_set_halt: 
>> ep2out: NOT stalled
>>
>> Sniffer shows me this completing, but I don't see IRQ for this.
>
> BTW, I just tested without $subject and it fails the same way. This is
> caused by something else. Can you rerun your bisect while I look at the
> problem here?

Okay, found it. This is caused by the ep_dequeue bug that I already
fixed. see [1] for that

[1] https://marc.info/?i=20170217105759.24356-1-felipe.ba...@linux.intel.com

-- 
balbi


signature.asc
Description: PGP signature

RE: [PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[Felipe Balbi]

Hi,

Felipe Balbi  writes:
>> Thinh Nguyen  writes:
  drivers/usb/dwc3/gadget.c | 5 +
  1 file changed, 5 insertions(+)
 
 diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
 index 6faf484e5dfc..0a664d8eba3f 100644
 --- a/drivers/usb/dwc3/gadget.c
 +++ b/drivers/usb/dwc3/gadget.c
 @@ -1379,6 +1379,9 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep
 *dep, int value, int protocol)
unsigned transfer_in_flight;
unsigned started;
 
 +  if (dep->flags & DWC3_EP_STALL)
 +  return 0;
 +
if (dep->number > 1)
trb = dwc3_ep_prev_trb(dep, dep->trb_enqueue);
else
 @@ -1400,6 +1403,8 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep
 *dep, int value, int protocol)
else
dep->flags |= DWC3_EP_STALL;
} else {
 +  if (!(dep->flags & DWC3_EP_STALL))
 +  return 0;
 
ret = dwc3_send_clear_stall_ep_cmd(dep);
if (ret)
 --
 2.11.0.295.gd7dffce1ce
 
 --
>>>
>>> I encounter an issue when I test mainline with USB 3 CV MSC test and
>>> bisected to this patch.
>>>
>>> Tester: USB 3 CV test (v2.1.3.0).
>>> Function Driver: f_mass_storage
>>> Reproducibility: always 
>>>
>>> The failure occurs under MSC Error Recovery Test (TD 1.4).
>>>
>>> Failure from the test sequence as follow (MSC compliance test spec) :
>>> 1. Place the device in the desired starting state.
>>> 2. Issue a Case 9 CBW (see Table 2) but with invalid signature of 
>>> 0xDEADBEEF.
>>> 3. Issue several In requests to the Bulk-Only Data Interface Bulk In 
>>> endpoint. Verify a
>>> STALL handshake is returned. If STALL handshake is not returned, skip to 
>>> step 11.
>>> 4. Issue a Get_Status(endpoint) request targeting the Bulk-Only Data 
>>> Interface bulk In
>>> endpoint. Verify that it completes normally, reports endpoint halt status.
>>>
>>>  ***After this point the device idles for 10 seconds and resets. ***
>>>
>>> Test fails.
>>
>> I've been looking at this and based on sniffer and dwc3 tracepoints, it
>> seems like dwc3 is behaving properly. The real issue seems to be that
>> g_mass_storage isn't queueing a new request to IN endpoint.
>>
>> I'll continue debugging this and try to find a solution that doesn't
>> involve reverting $subject.
>
> oh no, wait. ep2out misses XferInProgress:
>
> file-storage-1592  [000] d..1   152.809922: dwc3_ep_queue: ep2out: req 
> 88003cd6ee40 length 0/512 zsI ==> -115
> file-storage-1592  [000] d..1   152.809931: dwc3_prepare_trb: ep2out: 3/8 
> trb 88003a196050 buf 2d5e4000 size 512 ctrl 0819 
> (HlcS:sC:normal)
> file-storage-1592  [000] d..1   152.809942: dwc3_gadget_ep_cmd: ep2out: 
> cmd 'Update Transfer' [262151] params    --> status: 
> Successful
> file-storage-1592  [000]    152.809951: usb_ep_queue: ep2out: length 
> 0/512 sgs 0/0 stream 0 zsI status -115 --> 0
>  irq/34-dwc3-1593  [001] d..1   152.810212: dwc3_event: event (c040): 
> ep0out: Transfer Complete [Setup Phase]
>  irq/34-dwc3-1593  [001] d..1   152.810218: dwc3_ctrl_req: bRequestType 
> 02 bRequest 01 wValue  wIndex 0002 wLength 0
>  irq/34-dwc3-1593  [001] d..1   152.810228: __dwc3_gadget_ep_set_halt: 
> ep2out: NOT stalled
>
> Sniffer shows me this completing, but I don't see IRQ for this.

BTW, I just tested without $subject and it fails the same way. This is
caused by something else. Can you rerun your bisect while I look at the
problem here?

-- 
balbi


signature.asc
Description: PGP signature

RE: [PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[Felipe Balbi]

Hi Thinh,

Thinh Nguyen  writes:
>>  drivers/usb/dwc3/gadget.c | 5 +
>>  1 file changed, 5 insertions(+)
>> 
>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>> index 6faf484e5dfc..0a664d8eba3f 100644
>> --- a/drivers/usb/dwc3/gadget.c
>> +++ b/drivers/usb/dwc3/gadget.c
>> @@ -1379,6 +1379,9 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep
>> *dep, int value, int protocol)
>>  unsigned transfer_in_flight;
>>  unsigned started;
>> 
>> +if (dep->flags & DWC3_EP_STALL)
>> +return 0;
>> +
>>  if (dep->number > 1)
>>  trb = dwc3_ep_prev_trb(dep, dep->trb_enqueue);
>>  else
>> @@ -1400,6 +1403,8 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep
>> *dep, int value, int protocol)
>>  else
>>  dep->flags |= DWC3_EP_STALL;
>>  } else {
>> +if (!(dep->flags & DWC3_EP_STALL))
>> +return 0;
>> 
>>  ret = dwc3_send_clear_stall_ep_cmd(dep);
>>  if (ret)
>> --
>> 2.11.0.295.gd7dffce1ce
>> 
>> --
>
> I encounter an issue when I test mainline with USB 3 CV MSC test and
> bisected to this patch.
>
> Tester: USB 3 CV test (v2.1.3.0).
> Function Driver: f_mass_storage
> Reproducibility: always 
>
> The failure occurs under MSC Error Recovery Test (TD 1.4).
>
> Failure from the test sequence as follow (MSC compliance test spec) :
> 1. Place the device in the desired starting state.
> 2. Issue a Case 9 CBW (see Table 2) but with invalid signature of 0xDEADBEEF.
> 3. Issue several In requests to the Bulk-Only Data Interface Bulk In 
> endpoint. Verify a
> STALL handshake is returned. If STALL handshake is not returned, skip to step 
> 11.
> 4. Issue a Get_Status(endpoint) request targeting the Bulk-Only Data 
> Interface bulk In
> endpoint. Verify that it completes normally, reports endpoint halt status.
>
>  ***After this point the device idles for 10 seconds and resets. ***
>
> Test fails.

I've been looking at this and based on sniffer and dwc3 tracepoints, it
seems like dwc3 is behaving properly. The real issue seems to be that
g_mass_storage isn't queueing a new request to IN endpoint.

I'll continue debugging this and try to find a solution that doesn't
involve reverting $subject.

-- 
balbi


signature.asc
Description: PGP signature

RE: [PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[Felipe Balbi]

Felipe Balbi  writes:

> Hi Thinh,
>
> Thinh Nguyen  writes:
>>>  drivers/usb/dwc3/gadget.c | 5 +
>>>  1 file changed, 5 insertions(+)
>>> 
>>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>>> index 6faf484e5dfc..0a664d8eba3f 100644
>>> --- a/drivers/usb/dwc3/gadget.c
>>> +++ b/drivers/usb/dwc3/gadget.c
>>> @@ -1379,6 +1379,9 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep
>>> *dep, int value, int protocol)
>>> unsigned transfer_in_flight;
>>> unsigned started;
>>> 
>>> +   if (dep->flags & DWC3_EP_STALL)
>>> +   return 0;
>>> +
>>> if (dep->number > 1)
>>> trb = dwc3_ep_prev_trb(dep, dep->trb_enqueue);
>>> else
>>> @@ -1400,6 +1403,8 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep
>>> *dep, int value, int protocol)
>>> else
>>> dep->flags |= DWC3_EP_STALL;
>>> } else {
>>> +   if (!(dep->flags & DWC3_EP_STALL))
>>> +   return 0;
>>> 
>>> ret = dwc3_send_clear_stall_ep_cmd(dep);
>>> if (ret)
>>> --
>>> 2.11.0.295.gd7dffce1ce
>>> 
>>> --
>>
>> I encounter an issue when I test mainline with USB 3 CV MSC test and
>> bisected to this patch.
>>
>> Tester: USB 3 CV test (v2.1.3.0).
>> Function Driver: f_mass_storage
>> Reproducibility: always 
>>
>> The failure occurs under MSC Error Recovery Test (TD 1.4).
>>
>> Failure from the test sequence as follow (MSC compliance test spec) :
>> 1. Place the device in the desired starting state.
>> 2. Issue a Case 9 CBW (see Table 2) but with invalid signature of 0xDEADBEEF.
>> 3. Issue several In requests to the Bulk-Only Data Interface Bulk In 
>> endpoint. Verify a
>> STALL handshake is returned. If STALL handshake is not returned, skip to 
>> step 11.
>> 4. Issue a Get_Status(endpoint) request targeting the Bulk-Only Data 
>> Interface bulk In
>> endpoint. Verify that it completes normally, reports endpoint halt status.
>>
>>  ***After this point the device idles for 10 seconds and resets. ***
>>
>> Test fails.
>
> I've been looking at this and based on sniffer and dwc3 tracepoints, it
> seems like dwc3 is behaving properly. The real issue seems to be that
> g_mass_storage isn't queueing a new request to IN endpoint.
>
> I'll continue debugging this and try to find a solution that doesn't
> involve reverting $subject.

oh no, wait. ep2out misses XferInProgress:

file-storage-1592  [000] d..1   152.809922: dwc3_ep_queue: ep2out: req 
88003cd6ee40 length 0/512 zsI ==> -115
file-storage-1592  [000] d..1   152.809931: dwc3_prepare_trb: ep2out: 3/8 
trb 88003a196050 buf 2d5e4000 size 512 ctrl 0819 
(HlcS:sC:normal)
file-storage-1592  [000] d..1   152.809942: dwc3_gadget_ep_cmd: ep2out: cmd 
'Update Transfer' [262151] params    --> status: 
Successful
file-storage-1592  [000]    152.809951: usb_ep_queue: ep2out: length 
0/512 sgs 0/0 stream 0 zsI status -115 --> 0
 irq/34-dwc3-1593  [001] d..1   152.810212: dwc3_event: event (c040): 
ep0out: Transfer Complete [Setup Phase]
 irq/34-dwc3-1593  [001] d..1   152.810218: dwc3_ctrl_req: bRequestType 02 
bRequest 01 wValue  wIndex 0002 wLength 0
 irq/34-dwc3-1593  [001] d..1   152.810228: __dwc3_gadget_ep_set_halt: 
ep2out: NOT stalled

Sniffer shows me this completing, but I don't see IRQ for this.

-- 
balbi


signature.asc
Description: PGP signature

[PATCH] usb: dwc3: gadget: skip Set/Clear Halt when invalid

[Felipe Balbi]

At least macOS seems to be sending
ClearFeature(ENDPOINT_HALT) to endpoints which
aren't Halted. This makes DWC3's CLEARSTALL command
time out which causes several issues for the driver.

Instead, let's just return 0 and bail out early.

Cc: 
Signed-off-by: Felipe Balbi 
---

this falls into "has never worked before" category, so I'll be sending
it together with other patches for v4.11 merge window. Still, it's a
valid bug that's likely needed for stable trees.

 drivers/usb/dwc3/gadget.c | 5 +
 1 file changed, 5 insertions(+)

diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 6faf484e5dfc..0a664d8eba3f 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -1379,6 +1379,9 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, int 
value, int protocol)
unsigned transfer_in_flight;
unsigned started;
 
+   if (dep->flags & DWC3_EP_STALL)
+   return 0;
+
if (dep->number > 1)
trb = dwc3_ep_prev_trb(dep, dep->trb_enqueue);
else
@@ -1400,6 +1403,8 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, int 
value, int protocol)
else
dep->flags |= DWC3_EP_STALL;
} else {
+   if (!(dep->flags & DWC3_EP_STALL))
+   return 0;
 
ret = dwc3_send_clear_stall_ep_cmd(dep);
if (ret)
-- 
2.11.0.295.gd7dffce1ce

--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html