Re: [PATCH] mmc: renesas_sdhi_internal_dmac: Add MMC_CAP_ERASE to Gen3 SoCs

From: Eugeniu Rosca
Date: Fri Nov 15 2019 - 07:51:29 EST


Hello Yamada-san,

On Fri, Nov 15, 2019 at 10:27:25AM +0100, Ulf Hansson wrote:
> On Thu, 14 Nov 2019 at 23:07, Eugeniu Rosca <erosca@xxxxxxxxxxxxxx> wrote:
> >
> > Hi Ulf,
> >
> > On Thu, Nov 14, 2019 at 01:48:41PM +0100, Ulf Hansson wrote:
> >
> > [..]
> > >
> > > Let's first take a step back, because I don't know how the HW busy
> > > detection works for your controller.
> > >
> > > I have noticed there is TMIO_STAT_CMD_BUSY bit being set for some
> > > variants, which seems to cause renesas_sdhi_wait_idle() to loop for a
> > > pre-defined number of loops/timeout. This looks scary, but I can't
> > > tell if it's really a problem.
> > >
> > > BTW, do you know what TMIO_STAT_CMD_BUSY actually is monitoring?
> > >
> > > I have also noticed that MMC_CAP_WAIT_WHILE_BUSY isn't set for any of
> > > the renesas/tmio variant hosts. Is that simply because the HW doesn't
> > > support this? Or because implementation is missing?
> >
> > Hopefully Wolfram just addressed that?
> >
> > > If you want to run a test that stretches the behaviour on the timeout
> > > path, I would rather use an SD-card (the older the better). For eMMCs
> > > the erase likely translates to a trim/discard, which is far more
> > > quicker than a real erase - as is what happens on an old SD card.
> >
> > Running 'blkdiscard' with different SD cards on H3ULCB, I don't see any
> > signs of misbehavior:
> >
> > root@rcar-gen3:~# blkdiscard -V
> > blkdiscard from util-linux 2.32.1
> >
> > root@rcar-gen3:~# lsblk
> > NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
> > mmcblk0 179:0 0 59.2G 0 disk
> > mmcblk0boot0 179:8 0 4M 1 disk
> > mmcblk0boot1 179:16 0 4M 1 disk
> > mmcblk1 179:24 0 30G 0 disk
> >
> > # Erasing 32 GiB uSD Card
> > root@rcar-gen3:~# time blkdiscard -v /dev/mmcblk1
> > /dev/mmcblk1: Discarded 32227983360 bytes from the offset 0
> >
> > real 0m1.198s
> > user 0m0.001s
> > sys 0m0.122s
> >
> > # Erasing 64 GiB eMMC
> > root@rcar-gen3:~# time blkdiscard -v /dev/mmcblk0
> > /dev/mmcblk0: Discarded 63585648640 bytes from the offset 0
> >
> > real 0m8.703s
> > user 0m0.002s
> > sys 0m1.909s
> >
> > I guess that by decreasing below erase sizes, I could further increase
> > the execution time, but these sysfs properties are read-only:
> >
> > cat /sys/devices/platform/soc/ee100000.sd/mmc_host/mmc1/mmc1:59b4/preferred_erase_size
> > 4194304
> > cat /sys/devices/platform/soc/ee100000.sd/mmc_host/mmc1/mmc1:59b4/erase_size
> > 512
> >
>
> This test and due to the discussions with Wolfram and you in this
> thread, I would actually suggest that you enable MMC_CAP_ERASE for all
> tmio variants, rather than just for this particular one.
>
> In other words, set the cap in tmio_mmc_host_probe() should be fine,
> as it seems none of the tmio variants supports HW busy detection at
> this point.

Just for your information, following Ulf's suggestion, we are going to
enable MMC_CAP_ERASE in the TMIO mmc core driver, affecting UniPhier
SD/eMMC Host Controller. Hope to see your Ack/NAK on this in the
upcoming patch. TIA.

>
> Kind regards
> Uffe

--
Best Regards,
Eugeniu