Skip to contentSkip to conversation

usb: dwc3: gadget: Fix controller get stuck when kicking extra transfer in wrong case

PATCH In review Ran Wang 3 participants 4 messages Jan 7, 2020 – Jan 9, 2020

Subsystems (3)
lore
Share & tools
lore

Ran Wang

author
#0 lore raw

PATCH 1/1 usb: dwc3: gadget: Fix controller get stuck when kicking extra transfer in wrong case

According to original commit c96e6725db9d6a ("usb: dwc3: gadget: Correct the logic for queuing sgs"), we would only kick off another transfer in case of req->num_pending_sgs > 0.

However, current logic will do this as long as req->remaining > 0, this will include the case of non-sgs (both dwc3_gadget_ep_request_completed(req) and req->num_pending_sgs are 0) that we did not want to.

Without this fix, we observed dwc3 got stuck on Layerscape plaftorms (such as LS1088ARDB) when enabling gadget (mass storage function) as below:

[   27.923959] Mass Storage Function, version: 2009/09/11
[   27.929115] LUN: removable file: (no medium)
[   27.933432] LUN: file: /run/media/sda1/419/test
[   27.937963] Number of LUNs=1
[   27.941042] g_mass_storage gadget: Mass Storage Gadget, version: 2009/09/11
[   27.948019] g_mass_storage gadget: userspace failed to provide iSerialNumber
[   27.955069] g_mass_storage gadget: g_mass_storage ready
[   28.411188] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[   48.319766] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[   68.320794] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[   88.319898] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  108.320808] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  128.323419] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  148.320857] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  148.362023] g_mass_storage gadget: super-speed config #0: unconfigured
Fixes 8c7d4b7b3d43 ("usb: dwc3: gadget: Fix logical condition")
---
 drivers/usb/dwc3/gadget.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/usb/dwc3/gadget.c +1 -1
index 0c960a9..5b0f02f 100644
@@ -2491,7 +2491,7 @@ static int dwc3_gadget_ep_cleanup_completed_request(struct dwc3_ep *dep,
 
 	req->request.actual = req->request.length - req->remaining;
 
-	if (!dwc3_gadget_ep_request_completed(req) ||
+	if (!dwc3_gadget_ep_request_completed(req) &&
 			req->num_pending_sgs) {
 		__dwc3_gadget_kick_transfer(dep);
 		goto out;
Signature · 1 line
2.7.4
#1 lore raw

Hi,

Quoted 4 lines · Ran Wang

According to original commit c96e6725db9d6a ("usb: dwc3: gadget: Correct the logic for queuing sgs"), we would only kick off another transfer in case of req->num_pending_sgs > 0.

The commit 8c7d4b7b3d43 ("usb: dwc3: gadget: Fix logical condition") fixes the commit f38e35dd84e2 (usb: dwc3: gadget: split dwc3_gadget_ep_cleanup_completed_requests()).

Quoted 7 lines

However, current logic will do this as long as req->remaining > 0, this will include the case of non-sgs (both dwc3_gadget_ep_request_completed(req) and req->num_pending_sgs are 0) that we did not want to.

Without this fix, we observed dwc3 got stuck on Layerscape plaftorms (such as LS1088ARDB) when enabling gadget (mass storage function) as below:

Similar issue was reported by Thinh after my fix, and he has submitted a patch for the same. You can refer the discussion https://patchwork.kernel.org/patch/11292087/.

Quoted 30 lines
[   27.923959] Mass Storage Function, version: 2009/09/11
[   27.929115] LUN: removable file: (no medium)
[   27.933432] LUN: file: /run/media/sda1/419/test
[   27.937963] Number of LUNs=1
[   27.941042] g_mass_storage gadget: Mass Storage Gadget, version: 2009/09/11
[   27.948019] g_mass_storage gadget: userspace failed to provide iSerialNumber
[   27.955069] g_mass_storage gadget: g_mass_storage ready
[   28.411188] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[   48.319766] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[   68.320794] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[   88.319898] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  108.320808] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  128.323419] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  148.320857] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
[  148.362023] g_mass_storage gadget: super-speed config #0: unconfigured
Fixes 8c7d4b7b3d43 ("usb: dwc3: gadget: Fix logical condition")
---
 drivers/usb/dwc3/gadget.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/usb/dwc3/gadget.c +0 -0
index 0c960a9..5b0f02f 100644

drivers/usb/dwc3/gadget.c static int dwc3_gadget_ep_cleanup_completed_request(struct dwc3_ep *dep, · line 2497

Quoted 8 lines
drivers/usb/dwc3/gadget.c +1 -1
@@ -2491,7 +2491,7 @@ static int dwc3_gadget_ep_cleanup_completed_request(struct dwc3_ep *dep,
 
 	req->request.actual = req->request.length - req->remaining;
 
-	if (!dwc3_gadget_ep_request_completed(req) ||
+	if (!dwc3_gadget_ep_request_completed(req) &&
 			req->num_pending_sgs) {
 		__dwc3_gadget_kick_transfer(dep);
 		goto out;

Thanks & Regards,
Tejas Joglekar

author
#2 lore raw

Hi Tejas,

Quoted 10 lines · Tejas Joglekar

Hi,

Quoted 5 lines · Ran Wang

According to original commit c96e6725db9d6a ("usb: dwc3: gadget:
Correct the logic for queuing sgs"), we would only kick off another transfer in case of
req->num_pending_sgs > 0.

The commit 8c7d4b7b3d43 ("usb: dwc3: gadget: Fix logical condition") fixes the commit f38e35dd84e2 (usb: dwc3: gadget: split dwc3_gadget_ep_cleanup_completed_requests()).

Yes, actually I have tried to dig the history of this implementation as much as I can. and the commit I mentioned looks like the first one to add this function.

Quoted 9 lines
Quoted 9 lines

However, current logic will do this as long as req->remaining > 0, this will include the case of non-sgs (both dwc3_gadget_ep_request_completed(req) and req->num_pending_sgs are 0) that we did not want to.

Without this fix, we observed dwc3 got stuck on Layerscape plaftorms (such as
LS1088ARDB) when enabling gadget (mass storage function) as below:

Quoted 8 lines

Similar issue was reported by Thinh after my fix, and he has submitted a patch for the same. You can refer the discussion
https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpatch
work.kernel.org%2Fpatch%2F11292087%2F&data=02%7C01%7Cran.wan g_1%40nxp.com%7C97345f38c19849e5a82608d7935a8d31%7C686ea1d3bc2b 4c6fa92cd99c5c301635%7C0%7C0%7C637139889495749889&sdata=tcV6 FWjqvYYO7BCByACqAg%2FB%2B%2Fz3NC%2B20%2BIEVuyfJc4%3D&reser ved=0.

Thanks for this information. I will follow that thread.

Regards,
Ran

Quoted 49 lines
Quoted 5 lines
[   27.923959] Mass Storage Function, version: 2009/09/11
[   27.929115] LUN: removable file: (no medium)
[   27.933432] LUN: file: /run/media/sda1/419/test
[   27.937963] Number of LUNs=1
[   27.941042] g_mass_storage gadget: Mass Storage Gadget, version:

2009/09/11

Quoted 1 line
[   27.948019] g_mass_storage gadget: userspace failed to provide

iSerialNumber

Quoted 2 lines
[   27.955069] g_mass_storage gadget: g_mass_storage ready
[   28.411188] g_mass_storage gadget: super-speed config #1: Linux File-

Backed Storage

Quoted 1 line
[   48.319766] g_mass_storage gadget: super-speed config #1: Linux File-

Backed Storage

Quoted 1 line
[   68.320794] g_mass_storage gadget: super-speed config #1: Linux File-

Backed Storage

Quoted 1 line
[   88.319898] g_mass_storage gadget: super-speed config #1: Linux File-

Backed Storage

Quoted 27 lines
[  108.320808] g_mass_storage gadget: super-speed config #1: Linux

File-Backed Storage [ 128.323419] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage [ 148.320857] g_mass_storage
gadget: super-speed config #1: Linux File-Backed Storage [ 148.362023] g_mass_storage gadget: super-speed config #0: unconfigured

Fixes 8c7d4b7b3d43 ("usb: dwc3: gadget: Fix logical condition")
---
 drivers/usb/dwc3/gadget.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/usb/dwc3/gadget.c +1 -1
index 0c960a9..5b0f02f 100644
@@ -2491,7 +2491,7 @@ static int
 dwc3_gadget_ep_cleanup_completed_request(struct dwc3_ep *dep,
 
 	req->request.actual = req->request.length - req->remaining;
 
-	if (!dwc3_gadget_ep_request_completed(req) ||
+	if (!dwc3_gadget_ep_request_completed(req) &&
 			req->num_pending_sgs) {
 		__dwc3_gadget_kick_transfer(dep);
Quoted 2 lines

goto out;

Thanks & Regards,
Tejas Joglekar

#3 lore raw
In reply to Tejas Joglekar

Hi,

(Tejas, please break your lines at 80-columns. You shouldn't have to be reminded of this)

Quoted 14 lines · Tejas Joglekar

Hi,

Quoted 4 lines · Ran Wang

According to original commit c96e6725db9d6a ("usb: dwc3: gadget: Correct the logic for queuing sgs"), we would only kick off another transfer in case of req->num_pending_sgs > 0.

The commit 8c7d4b7b3d43 ("usb: dwc3: gadget: Fix logical condition") fixes the commit f38e35dd84e2 (usb: dwc3: gadget: split dwc3_gadget_ep_cleanup_completed_requests()).

Quoted 4 lines

However, current logic will do this as long as req->remaining > 0, this will include the case of non-sgs (both dwc3_gadget_ep_request_completed(req) and req->num_pending_sgs are 0) that we did not want to.

Quoted 6 lines
Quoted 3 lines

Without this fix, we observed dwc3 got stuck on Layerscape plaftorms (such as LS1088ARDB) when enabling gadget (mass storage function) as below:

Similar issue was reported by Thinh after my fix, and he has submitted a patch for the same. You can refer the discussion
https://patchwork.kernel.org/patch/11292087/.

Based on this, I'm assuming we don't need this patch.

Signature · 1 line
balbi