From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by mx.groups.io with SMTP id smtpd.web11.14042.1582128326991781892 for ; Wed, 19 Feb 2020 08:05:27 -0800 Authentication-Results: mx.groups.io; dkim=missing; spf=pass (domain: intel.com, ip: 192.55.52.88, mailfrom: mateusz.albecki@intel.com) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga003.jf.intel.com ([10.7.209.27]) by fmsmga101.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 19 Feb 2020 08:05:14 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.70,461,1574150400"; d="scan'208";a="235932974" Received: from gklab-27-32.ger.corp.intel.com ([10.102.28.45]) by orsmga003.jf.intel.com with ESMTP; 19 Feb 2020 08:05:12 -0800 From: "Albecki, Mateusz" To: devel@edk2.groups.io Cc: Mateusz Albecki , Hao A Wu , Marcin Wojtas , Zhichao Gao , Liming Gao Subject: [PATCHv2 0/4] MdeModulePkg/SdMmcPciHcDxe: Refactor command processing Date: Wed, 19 Feb 2020 17:05:04 +0100 Message-Id: <20200219160508.5856-1-mateusz.albecki@intel.com> X-Mailer: git-send-email 2.14.1.windows.1 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 Cc: Marcin Wojtas Cc: Zhichao Gao Cc: Liming Gao 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.