The sg_miter used to loop over the returned sglist from a
transfer in the esdhc subdriver for SDHCI can be called
from atomic context so the miter needs to be atomic.
sdhci_request_done() is always called from process context,
either as a work or as part of the threaded interrupt handler,
but the one case when we are actually calling .request_done()
from an atomic context is in sdhci_irq().
Fix this by flaggig the miter atomic so we always use
kmap_atomic().
Fixes: e8a167b84886 ("mmc: sdhci-esdhc-mcf: Use sg_miter for swapping")
Reported-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
Signed-off-by: Linus Walleij <linus.walleij@xxxxxxxxxx>
---
Changes in v2:
- Do the simpler fix and just use an atomic iterator, it
will work fine since esdhc_mcf_buffer_swap32() doesn't sleep.
- Link to v1: https://lore.kernel.org/r/20240222-fix-sdhci-esdhc-mcf-v1-1-fb87e04ca575@xxxxxxxxxx
---
drivers/mmc/host/sdhci-esdhc-mcf.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/mmc/host/sdhci-esdhc-mcf.c b/drivers/mmc/host/sdhci-esdhc-mcf.c
index 1909a11fd065..c97363e2d86c 100644
--- a/drivers/mmc/host/sdhci-esdhc-mcf.c
+++ b/drivers/mmc/host/sdhci-esdhc-mcf.c
@@ -313,7 +313,7 @@ static void esdhc_mcf_request_done(struct sdhci_host *host,
* transfer endiannes. A swap after the transfer is needed.
*/
sg_miter_start(&sgm, mrq->data->sg, mrq->data->sg_len,
- SG_MITER_TO_SG | SG_MITER_FROM_SG);
+ SG_MITER_ATOMIC | SG_MITER_TO_SG | SG_MITER_FROM_SG);
while (sg_miter_next(&sgm)) {
buffer = sgm.addr;
esdhc_mcf_buffer_swap32(buffer, sgm.length);
---
base-commit: 20af1ca418d2c0b11bc2a1fe8c0c88f67bcc2a7e
change-id: 20240228-fix-sdhci-esdhc-mcf-2-539cf338c3a3
Best regards,
--
Linus Walleij <linus.walleij@xxxxxxxxxx>
