Hi Adrian, On Mon, Nov 19, 2018 at 09:37:40AM +0200, Adrian Hunter wrote: > On 16/11/18 6:57 PM, Anisse Astier wrote: > > Hi Adrian, > > > > On Fri, Nov 16, 2018 at 04:01:36PM +0200, Adrian Hunter wrote: > >> On 16/11/18 11:50 AM, Adrian Hunter wrote: > >>> On 16/11/18 11:37 AM, Anisse Astier wrote: > >>>> Hi Adrian, > >>>> > >>>> On Thu, Nov 08, 2018 at 05:43:32PM +0000, Hunter, Adrian wrote: > >>>>> [snip] > >>>>> I found some time today. It seems that the ACPI _PS3 method is failing to save the tuning value. > >>>>> That results in a CRC error, but the driver does not recover very gracefully. > >>>>> > >>>>> I will look at making a patch for the driver recovery, but also add some diagnostics prints to the ACPI DSDT to try to figure out what is going wrong there. > >>>>> > >>>>> > >>>> > >>>> Have you had the time to look at what diagnostics could be added ? > >>> > >>> I have a patch, but I am still testing it. > >>> > >> > >> Please try the attached patch > > > >> From 5c33302367e7852aa0a723e311b1749bea59f055 Mon Sep 17 00:00:00 2001 > >> From: Adrian Hunter <adrian.hunter@xxxxxxxxx> > >> Date: Fri, 16 Nov 2018 14:02:12 +0200 > >> Subject: [PATCH] mmc: sdhci-pci: Workaround GLK firmware failing to restore > >> the tuning value > >> > >> GLK firmware can indicate that the tuning value will be restored after > >> runtime suspend, but not actually do that. Add a workaround that detects > >> such cases, and lets the driver do re-tuning instead. > >> > > > > This looks like it solves the issue. Please find the boot log here: > > > > https://anisse.astier.eu/static/dmesg-sdhci-glk-workaround.txt.xz > > > > We still have I/O errors I don't understand, but no more long CQE > > timeouts. > > Please try this patch instead. > From ff445c899af814f0623ef191213b9d7ddb51aa69 Mon Sep 17 00:00:00 2001 > From: Adrian Hunter <adrian.hunter@xxxxxxxxx> > Date: Fri, 16 Nov 2018 14:02:12 +0200 > Subject: [PATCH V2] mmc: sdhci-pci: Workaround GLK firmware failing to restore > the tuning value > > GLK firmware can indicate that the tuning value will be restored after > runtime suspend, but not actually do that. Add a workaround that detects > such cases, and lets the driver do re-tuning instead. > > Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx> Here is the boot log for the v2 of the patch: https://anisse.astier.eu/static/dmesg-sdhci-glk-workaround-v2.txt.xz There are no more I/O errors. Regards, Anisse
