public inbox for devel@edk2.groups.io
 help / color / mirror / Atom feed
* [PATCHv2 0/4] MdeModulePkg/SdMmcPciHcDxe: Refactor command processing
@ 2020-02-19 16:05 Albecki, Mateusz
  2020-02-19 16:05 ` [PATCHv2 1/4] MdeModulePkg/SdMmcPciHcDxe: Enhance driver traces Albecki, Mateusz
                   ` (4 more replies)
  0 siblings, 5 replies; 8+ messages in thread
From: Albecki, Mateusz @ 2020-02-19 16:05 UTC (permalink / raw)
  To: devel; +Cc: Mateusz Albecki, Hao A Wu, Marcin Wojtas, Zhichao Gao, Liming Gao

This patch series aims to refactor command processing to achieve following

- Trace the failing TRB packets to see what commands are failing and for what reasons
- Get the response data even if data transfer timed out to allow easier debugging
- Fix the PIO mode which is currently completely broken.

Changes in v2:
- Moved verbose packet prints after the command is finished to capture the successfull command response
- Fixed the debug prints
- PIO data will be moved with width matching the alignment of the block size. For majority of transfers that means UINT32 width.

Tests performed:
- Each patch in the series has passed boot from eMMC with ADMAv3 data transfer mode
- SDMA based boot has been tested with the full patch series
- PIO based boot has been tested with the full patch series
- PIO based data transfer has been additionally tested by creating and modyfing a file in EFI shell
- Tested async PIO transfer - results below

Async test results:
I've tested a simple async write and then readback from the eMMMC device using block IO v2. I have observed
that while the requests are processed correctly by the eMMC driver as soon as they finish platform will sometimes
mibehave. I've tested async without my changes and the strange behavior reproduces. Right now I am suspecting there
is some problem either in the EDK2 core that performs async or in the platform specific code. It is also possible that
I coded the Async BlockIo incorrectly although the test code was rather simple. Additionally I was able
to observe that on many eMMC controllers PIO based data transfer is broken. I was only able to find one platform
that supported PIO. Detailed observation below

Platform 1 (PIO working).
- Test code was able to perform async write to eMMC LBA 0
- As soon as the callback is called CPU exception happens
- After reboot test code was able to perform async read
- As soon as the callback is called CPU exception happens
- After reboot sync read was able to confirm that data matches what was written by async write

Platform2 (PIO is returning trash data - all 0xAF)
- Test code was able to perform async write(although it didn't realyl came through to the device)
- After write finished test code was able to perform async read(again all data was 0xAF but the logic in the driver works)

Platform2 (again PIO is returning trash data)
- Test code scheduled 5 async writes. 2 writes finished and after that CPU exception was signaled.

Platform3(also trash data from PIO)
- Test code scheduled one async write as soon as it was done platform rebooted

I didn't want to spend any more time debugging this issue as I think it would turn into the platform debug and what I
observed gave me some confidence that PIO in async is generally working.

All tests were performed with eMMC in HS400 @200MHz clock frequency.

For easier review & integration patch has been pushed here:
Whole series: https://github.com/malbecki/edk2/tree/emmc_transfer_refactor
Whole series + SDMA force code(test 3): https://github.com/malbecki/edk2/tree/emmc_transfer_refactor_force_sdma
Whole series + PIO force code(test 4): https://github.com/malbecki/edk2/tree/emmc_transfer_refactor_force_pio

Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Marcin Wojtas <mw@semihalf.com>
Cc: Zhichao Gao <zhichao.gao@intel.com>
Cc: Liming Gao <liming.gao@intel.com>

Mateusz Albecki (4):
  MdeModulePkg/SdMmcPciHcDxe: Enhance driver traces
  MdeModulePkg/SdMmcPciHcDxe: Read response on command completion
  MdeModulePkg/SdMmcPciHcDxe: Refactor data transfer completion
  MdeModulePkg/SdMmcPciHcDxe: Fix PIO transfer mode

 MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHcDxe.h |   4 +
 MdeModulePkg/Bus/Pci/SdMmcPciHcDxe/SdMmcPciHci.c   | 521 ++++++++++++++++-----
 2 files changed, 417 insertions(+), 108 deletions(-)

-- 
2.14.1.windows.1

--------------------------------------------------------------------

Intel Technology Poland sp. z o.o.
ul. Slowackiego 173 | 80-298 Gdansk | Sad Rejonowy Gdansk Polnoc | VII Wydzial Gospodarczy Krajowego Rejestru Sadowego - KRS 101882 | NIP 957-07-52-316 | Kapital zakladowy 200.000 PLN.

Ta wiadomosc wraz z zalacznikami jest przeznaczona dla okreslonego adresata i moze zawierac informacje poufne. W razie przypadkowego otrzymania tej wiadomosci, prosimy o powiadomienie nadawcy oraz trwale jej usuniecie; jakiekolwiek
przegladanie lub rozpowszechnianie jest zabronione.
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). If you are not the intended recipient, please contact the sender and delete all copies; any review or distribution by
others is strictly prohibited.


^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2020-02-21 14:31 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-02-19 16:05 [PATCHv2 0/4] MdeModulePkg/SdMmcPciHcDxe: Refactor command processing Albecki, Mateusz
2020-02-19 16:05 ` [PATCHv2 1/4] MdeModulePkg/SdMmcPciHcDxe: Enhance driver traces Albecki, Mateusz
2020-02-19 16:05 ` [PATCHv2 2/4] MdeModulePkg/SdMmcPciHcDxe: Read response on command completion Albecki, Mateusz
2020-02-19 16:05 ` [PATCHv2 3/4] MdeModulePkg/SdMmcPciHcDxe: Refactor data transfer completion Albecki, Mateusz
2020-02-19 16:05 ` [PATCHv2 4/4] MdeModulePkg/SdMmcPciHcDxe: Fix PIO transfer mode Albecki, Mateusz
2020-02-20  0:39 ` [PATCHv2 0/4] MdeModulePkg/SdMmcPciHcDxe: Refactor command processing Wu, Hao A
2020-02-20 12:35   ` Albecki, Mateusz
     [not found]   ` <15F51C7BE6896999.18319@groups.io>
2020-02-21 14:30     ` [edk2-devel] " Albecki, Mateusz

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox