On Mon, Aug 08, 2022 at 03:43:17PM +0200, Ahmad Fatoum wrote:
> Hello Sascha,
>
> On 08.08.22 15:40, Sascha Hauer wrote:
> > On Mon, Jul 25, 2022 at 01:58:56PM +0200, Ahmad Fatoum wrote:
> >> In remote labs co-located with other hardware, we've observed card
> >> detect levers of different boards to sporadically fail to detect
> >> the card, e.g. because the cable on the usbsdmux was yanked around
> >> by accident. When this happens, barebox usually boots up normally as
> >> the card detect is ignored and then Linux waits indefinitely for
> >> the card-detect to turn active. Add a new config option that can be
> >> enabled to avoid these issues altogether.
> >>
> >> Signed-off-by: Ahmad Fatoum <a.fatoum@xxxxxxxxxxxxxx>
> >> ---
> >> v1 -> v2:
> >> - use early exit instead of if-else
> >> - don't fail mci_card_probe if of_register_fixup fails
> >> ---
> >> drivers/mci/Kconfig | 15 +++++++++++++++
> >> drivers/mci/mci-core.c | 36 ++++++++++++++++++++++++++++++++----
> >> 2 files changed, 47 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/drivers/mci/Kconfig b/drivers/mci/Kconfig
> >> index 21d53c0c3f0b..651e59259790 100644
> >> --- a/drivers/mci/Kconfig
> >> +++ b/drivers/mci/Kconfig
> >> @@ -56,6 +56,21 @@ config MCI_MMC_GPP_PARTITIONS
> >> Note: by default, 'MMC' devices have no 'general purpose partitions',
> >> it requires a special one-time configuration step to enable them.
> >>
> >> +config MCI_BROKEN_CD
> >> + bool "ignore card-detect pin on boot and in OS"
> >> + help
> >> + Say 'y' here to have barebox unconditionally ignore the
> >> + card-detect pin for its own operation and manipulate the
> >> + kernel DT, so all detected MCI cards are polled instead
> >> + of expecting the card detect lever to behave correctly.
> >> + If you need more fine grained control use of_property
> >> + in an init script:
> >> +
> >> + of_property -fd mmc0 cd-gpios
> >> + of_property -fs mmc0 broken-cd
> >> +
> >> + If unsure, say 'n' here.
> >> +
> >> comment "--- MCI host drivers ---"
> >>
> >> config MCI_DW
> >> diff --git a/drivers/mci/mci-core.c b/drivers/mci/mci-core.c
> >> index 82e2f82f5377..4db2253c3a80 100644
> >> --- a/drivers/mci/mci-core.c
> >> +++ b/drivers/mci/mci-core.c
> >> @@ -1727,6 +1727,27 @@ static int mci_register_partition(struct mci_part *part)
> >> return 0;
> >> }
> >>
> >> +static int of_broken_cd_fixup(struct device_node *root, void *ctx)
> >> +{
> >> + struct device_d *hw_dev = ctx;
> >> + struct device_node *np;
> >> + char *name;
> >> +
> >> + name = of_get_reproducible_name(hw_dev->device_node);
> >> + np = of_find_node_by_reproducible_name(root, name);
> >> + free(name);
> >> + if (!np) {
> >> + dev_warn(hw_dev, "Cannot find nodepath %s, cannot fixup\n",
> >> + hw_dev->device_node->full_name);
> >> + return -EINVAL;
> >> + }
> >> +
> >> + of_property_write_bool(np, "cd-gpios", false);
> >> + of_property_write_bool(np, "broken-cd", true);
> >> +
> >> + return 0;
> >> +}
> >> +
> >> /**
> >> * Probe an MCI card at the given host interface
> >> * @param mci MCI device instance
> >> @@ -1737,10 +1758,13 @@ static int mci_card_probe(struct mci *mci)
> >> struct mci_host *host = mci->host;
> >> int i, rc, disknum, ret;
> >>
> >> - if (host->card_present && !host->card_present(host) &&
> >> - !host->non_removable) {
> >> - dev_err(&mci->dev, "no card inserted\n");
> >> - return -ENODEV;
> >> + if (host->card_present && !host->card_present(host) && !host->non_removable) {
> >> + if (!IS_ENABLED(CONFIG_MCI_BROKEN_CD)) {
> >> + dev_err(&mci->dev, "no card inserted\n");
> >> + return -ENODEV;
> >> + }
> >> +
> >> + dev_info(&mci->dev, "no card inserted (ignoring)\n");
> >> }
> >>
> >> ret = regulator_enable(host->supply);
> >> @@ -1818,6 +1842,10 @@ static int mci_card_probe(struct mci *mci)
> >> }
> >> }
> >>
> >> + if (IS_ENABLED(CONFIG_MCI_BROKEN_CD) && !host->no_sd &&
> >> + dev_of_node(host->hw_dev))
> >> + of_register_fixup(of_broken_cd_fixup, host->hw_dev);
> >
> > This fixup is only registered when the card is probed, i.e. when it's
> > used in barebox. Shouldn't it be registered in mci_register() instead?
>
> This is intentional. We only know that cards barebox successfully probed
> are indeed available.
I don't get this argument. If card detect is broken then why can't we
let Linux go through the same steps as barebox when barebox hasn't done
so?
Letting the card work in Linux only when it was previously used in barebox
is surprising to me.
Sascha
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
