On Wed, Feb 21, 2024 at 10:23 PM Linus Walleij <linus.walleij@xxxxxxxxxx> wrote:
> The introduction of sg_miter was a bit sloppy as it didn't
> exactly mimic the semantics of the old code on multiblock reads
> and writes: these like you to:
>
> - Advance to the first sglist entry *before* starting to read
> any blocks from the card.
>
> - Advance and check availability of the next entry *right after*
> processing one block.
>
> Not checking if we have more sglist entries right after
> reading a block will lead to this not being checked until we
> return to the callback to read out more blocks, i.e. until the
> next interrupt arrives. Since the last block is the last one
> (no more data will arrive) there will not be a next interrupt,
> and we will be waiting forever resulting in a timeout for
> command 18 when reading multiple blocks.
>
> The same bug was fixed also in the writing of multiple blocks.
>
> Reported-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> Fixes: 27b57277d9ba ("mmc: sh_mmcif: Use sg_miter for PIO")
> Signed-off-by: Linus Walleij <linus.walleij@xxxxxxxxxx>
Thanks, eMMC works again (tested on APE6EVM and KZM9-GT-Dual).
Tested-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
