Julius Werner submitted this change.
sc7180: Fix for hang during DMA transfer in SPI-NOR flash driver
Transfer sequence used by SPI-Flash application present in CB/DC.
1. Assert CS through GPIO
2. Data transfer through QSPI (involves construction of command
descriptor for multiple read/write transfers)
3. De-assert CS through GPIO.
With above sequence, in DMA mode we dont have the support for read
transfers that are not preceded by write transfer in QSPI controller.
Ex: "write read read read" sequence results in hang during DMA transfer,
where as "write read write read" sequence has no issue.
As we have application controlling CS through GPIO, we are making
fragment bit "set" for all transfers, which keeps CS in asserted
state although the ideal way to operate CS is through QSPI controller.
Change-Id: Ia45ab793ad05861b88e99a320b1ee9f10707def7
Signed-off-by: satya priya <skakit@codeaurora.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/39807
Reviewed-by: Julius Werner <jwerner@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
---
M src/soc/qualcomm/sc7180/qspi.c
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/src/soc/qualcomm/sc7180/qspi.c b/src/soc/qualcomm/sc7180/qspi.c
index 10038e4..675641b 100644
--- a/src/soc/qualcomm/sc7180/qspi.c
+++ b/src/soc/qualcomm/sc7180/qspi.c
@@ -118,17 +118,20 @@
next->direction = MASTER_READ;
next->multi_io_mode = 0;
next->reserved1 = 0;
- next->fragment = 0;
+ /*
+ * QSPI controller doesn't support transfer starts with read segment.
+ * So to support read transfers that are not preceded by write, set
+ * transfer fragment bit = 1
+ */
+ next->fragment = 1;
next->reserved2 = 0;
next->length = 0;
next->bounce_src = 0;
next->bounce_dst = 0;
next->bounce_length = 0;
- if (current) {
+ if (current)
current->next_descriptor = (uint32_t)(uintptr_t) next;
- current->fragment = 1;
- }
return next;
}
To view, visit change 39807. To unsubscribe, or for help writing mail filters, visit settings.