Re: [PATCH] MdeModulePkg/UfsPassThruDxe: Improve Error handling of Ufs Pass Thru driver


Wu, Hao A
 

-----Original Message-----
From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Wu, Hao
A
Sent: Monday, February 22, 2021 4:38 PM
To: Bandaru, Purna Chandra Rao <purna.chandra.rao.bandaru@intel.com>;
devel@edk2.groups.io
Cc: Albecki, Mateusz <mateusz.albecki@intel.com>; Ni, Ray
<ray.ni@intel.com>
Subject: Re: [edk2-devel] [PATCH] MdeModulePkg/UfsPassThruDxe:
Improve Error handling of Ufs Pass Thru driver

-----Original Message-----
From: Bandaru, Purna Chandra Rao
<purna.chandra.rao.bandaru@intel.com>
Sent: Wednesday, February 17, 2021 5:02 PM
To: devel@edk2.groups.io
Cc: Bandaru, Purna Chandra Rao <purna.chandra.rao.bandaru@intel.com>;
Albecki, Mateusz <mateusz.albecki@intel.com>; Ni, Ray
<ray.ni@intel.com>; Wu, Hao A <hao.a.wu@intel.com>
Subject: [PATCH] MdeModulePkg/UfsPassThruDxe: Improve Error handling
of Ufs Pass Thru driver

From: Bandaru <purna.chandra.rao.bandaru@intel.com>

https://bugzilla.tianocore.org/show_bug.cgi?id=3217

Following is the brief description of the changes
1) There are cards that can take upto 600ms for Init and hence increase
the time out for fDeviceInit polling loop.
2) Add UFS host conctroller reset in the last retry of Link start up.
3) Retry sending NOP OUT command upto 10 times

Hello Bandaru,

Could you help to break this patch into a 3-patch series in V2?
With each patch handling just one of the above 3 improvements mentioned.

For improvement 2) above, I do not see such UFS host controller re-enabling
process being mentioned in UFSHCI 3.0 spec section 7.1.1.
Is this process being documented somewhere else in the spec or suggested
by device vender?

Sorry for missing one comment.
Could you help to add the information on what kind of tests have been performed for the code changes?

Thanks in advance.

Best Regards,
Hao Wu



More inline comments below:



Signed-off-by: Bandaru <purna.chandra.rao.bandaru@intel.com>
Cc: Mateusz Albecki <mateusz.albecki@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Hao A Wu <hao.a.wu@intel.com>

Change-Id: I6c0dbc1c147487e51f0ed5f2425957ae089b0160
---
MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThru.c | 26
+++++++++++++++++++++-----
MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThruHci.c | 18
++++++++++++------
2 files changed, 33 insertions(+), 11 deletions(-)

diff --git a/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThru.c
b/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThru.c
index 9768c2e6fb..89048745be 100644
--- a/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThru.c
+++ b/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThru.c
@@ -1,6 +1,6 @@
/** @file

- Copyright (c) 2014 - 2019, Intel Corporation. All rights
reserved.<BR>
+ Copyright (c) 2014 - 2021, Intel Corporation. All rights
+ reserved.<BR>
Copyright (c) Microsoft Corporation.<BR>
SPDX-License-Identifier: BSD-2-Clause-Patent

@@ -749,7 +749,7 @@ UfsFinishDeviceInitialization ( {
EFI_STATUS Status;
UINT8 DeviceInitStatus;
- UINT8 Timeout;
+ UINT16 Timeout;

DeviceInitStatus = 0xFF;

@@ -761,17 +761,23 @@ UfsFinishDeviceInitialization (
return Status;
}

- Timeout = 5;
+ Timeout = 6000; //There are cards that can take upto 600ms.

Please help to add a macro in file UfsPassThru.h:
#define UFS_INIT_COMPLETION_TIMEOUT 6000 And use the macro here.

Also a minor comment, could you help to use the below comment format?
//
// There are UFS devices that can take up to 600ms to clear the fDeviceInit
flag // Timeout = UFS_INIT_COMPLETION_TIMEOUT;


do {
+ MicroSecondDelay (100); //Give 100 us and then start polling.

For the above delay movement, do you observe any side effect for the origin
code?
If not, I prefer to leave the origin behavior:
do {
UfsReadFlag();
...
MicroSecondDelay (1);
} while (...)
since doing so will have the least performance penalty for devices that
respond fast.


Status = UfsReadFlag (Private, UfsFlagDevInit, &DeviceInitStatus);
if (EFI_ERROR (Status)) {
return Status;
}
- MicroSecondDelay (1);
Timeout--;
} while (DeviceInitStatus != 0 && Timeout != 0);

+ if (Timeout == 0) {
+ DEBUG ((DEBUG_ERROR, "UfsFinishDeviceInitialization
DeviceInitStatus=%x EFI_TIMEOUT \n", DeviceInitStatus));
+ return EFI_TIMEOUT;
+ } else {
+ DEBUG ((DEBUG_INFO, "UfsFinishDeviceInitialization Timeout
+ left=%x EFI_SUCCESS \n", Timeout));
return EFI_SUCCESS;

Please help to add two spaces for text alignment in the above line.


+ }
}

/**
@@ -905,9 +911,19 @@ UfsPassThruDriverBindingStart (
// At the end of the UFS Interconnect Layer initialization on both
host and device side,
// the host shall send a NOP OUT UPIU to verify that the device UTP
Layer is ready.
//

For the NOP OUT - NOP IN improvement, could you help to provide more
information on what is the current issue for some devices?
Is it a timeout happened for:
Status = UfsWaitMemSet (Private, UFS_HC_UTRLDBR_OFFSET, BIT0 << Slot,
0, UFS_TIMEOUT); (If so, have you tried increasing the last parameter like
'10*UFS_TIMEOUT'?) Or the case is that NopInUpiu->Resp has a non-zero
value?

I found that in the UFS 3.0 spec:
|> For some implementations, the device UTP layer may not be initialized
|> yet, therefore the device may not respond promptly to NOP OUT UPIU
|> sending NOP IN UPIU.
|> The host waits until it receives the NOP IN UPIU from the device...
And there is no mention for the retry scheme.


+ for (Index = 10; Index > 0; Index--) {
Status = UfsExecNopCmds (Private);
if (EFI_ERROR (Status)) {
- DEBUG ((DEBUG_ERROR, "Ufs Sending NOP IN command Error, Status
= %r\n", Status));
+ DEBUG ((DEBUG_ERROR, "Ufs Sending NOP IN command Error, Index
= %x Status = %r\n", Index, Status));
+ MicroSecondDelay (100); //100 us
+ continue;
+ } else {
+ DEBUG ((DEBUG_INFO, "Ufs Sent NOP OUT successfully and received
NOP IN, Status = %r\n", Status));
+ break;
+ }
+ }
+ if (!Index) {
+ DEBUG ((DEBUG_INFO, "NOP OUT failed all the 10 times Status =
+ %r\n", Status));
goto Error;
}

diff --git a/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThruHci.c
b/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThruHci.c
index 0b1030ab47..4fa5689196 100644
--- a/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThruHci.c
+++ b/MdeModulePkg/Bus/Ufs/UfsPassThruDxe/UfsPassThruHci.c
@@ -2,7 +2,7 @@
UfsPassThruDxe driver is used to produce EFI_EXT_SCSI_PASS_THRU
protocol interface
for upper layer application to execute UFS-supported SCSI cmds.

- Copyright (c) 2014 - 2019, Intel Corporation. All rights
reserved.<BR>
+ Copyright (c) 2014 - 2021, Intel Corporation. All rights
+ reserved.<BR>
Copyright (c) Microsoft Corporation.<BR>
SPDX-License-Identifier: BSD-2-Clause-Patent

@@ -1929,17 +1929,15 @@ UfsDeviceDetection (

//
// Start UFS device detection.
- // Try up to 3 times for establishing data link with device.
+ // Try up to 4 times for establishing data link with device.
//
- for (Retry = 0; Retry < 3; Retry++) {
+ for (Retry = 0; Retry < 4; Retry++) {

Please introduce a macro in file UfsPassThru.h:
#define UFS_LINK_STARTUP_RETRIES 4
And use the macro here.

Also, is it necessary to increase the retry number by 1?
Or the device can be successfully brought up by adding a host controller re-
enabling?


LinkStartupCommand.Opcode = UfsUicDmeLinkStartup;
LinkStartupCommand.Arg1 = 0;
LinkStartupCommand.Arg2 = 0;
LinkStartupCommand.Arg3 = 0;
Status = UfsExecUicCommands (Private, &LinkStartupCommand);
- if (EFI_ERROR (Status)) {
- return EFI_DEVICE_ERROR;
- }

Will the DME_LINKSTARTUP command execution fail at first and then
succeed after retry?
If not, I prefer to keep the origin code logic to return error status directly.


+ if (!EFI_ERROR (Status)) {

Status = UfsMmioRead32 (Private, UFS_HC_STATUS_OFFSET, &Data);
if (EFI_ERROR (Status)) {
@@ -1960,6 +1958,14 @@ UfsDeviceDetection (
}
}
return EFI_SUCCESS;
+ }
+ }
+ if (Retry == 2) {

Please help to update to:
if (Retry == UFS_LINK_STARTUP_RETRIES - 1) {

And add comments like:
//
// Try re-enabling the UFS host controller in the last retry attempt //


Best Regards,
Hao Wu


+ Status = UfsEnableHostController (Private);
+ if (EFI_ERROR (Status)) {
+ DEBUG ((DEBUG_ERROR, "UfsDeviceDetection: Enable Host
Controller
Fails, Status = %r\n", Status));
+ return Status;
+ }
}
}

--
2.16.2.windows.1



Join devel@edk2.groups.io to automatically receive all group messages.