Hi Wolfram-san, > From: Wolfram Sang, Sent: Wednesday, November 25, 2020 5:56 PM <snip> > > The following commit [1] is a BSP local patch though, > > we need to set -EILSEQ to retune a card for R-Car Gen3 [2] > > by MMC core driver [3]. > > So, if there is a non-removable eMMC or a SD card inserted, then we need > to EILSEQ to enforce a retune. Otherwise it is a data timeout. Is my > understanding correct? Yes. > I wonder, though, if "Gen3" is a complete description? There are SDHI > instances on Gen2 which can also do SDR104. Won't they need the same > treatment? Then we could say that every SDHI which has an SCC will need > this treatment. Hmm, since Gen2 datasheet didn't update the flowchart so that I thought we need special handling for "Gen3". But, I'll ask hardware team whether my thought is correct or not. > > - The patch also change the tmio_mmc_cmd_irq() when CMDTIMEOUT happens for R-Car Gen3. > > But, for upstream, we should make a separated patch for it. > > I am sorry. I don't fully understand. Why does the change to > tmio_mmc_cmd_irq() need a seperate patch? The patch subject is "Add data timeout error detection". That's why I thought so. > > - These "for R-Car Gen3" means I'm thinking we need additional condition: > > 1) to set -EILSRQ or -ETIMEDOUT for R-Car Gen3 > > 2) to set -ETIMEDOUT anyway for other SoCs. > > # These are complex conditions a little though... > > Well, from what I understood this sounds not too hard. Let's hope I just > got it correctly :) I got it :) > However, there is something in this patch which makes mmc_test #15 work, > though. We still want this in this series, or do you think it is better > to move it to a seperate series? Thank you for the comments. I realized mmc_test #15 failed if tmio driver returned -EILSEQ when I applied the BSP patch as the following: --- [ 206.016193] mmc0: Starting tests of card mmc0:0001... [ 206.022708] mmc0: Test case 15. Proper xfer_size at write (start failure)... [ 206.854082] mmc0: Result: ERROR (-84) [ 206.858632] mmc0: Tests completed. --- So, I'm thinking retuning -EILSEQ in cmd/data timeout should be a separate series now. # Perhaps we need other way to follow the retuning sequence for R-Car Gen3... Best regards, Yoshihiro Shimoda
