drivers: flash: fix RWW issues in flash_mcux_flexspi_hyperflash #88469
Conversation
github-actions
bot
added
area: Flash
platform: NXP Drivers
github-actions
bot
requested review from
dbaluta,
de-nordic,
decsny,
dleach02,
manuargue,
mmahadevan108,
Raymond0225 and
rghaddab
danieldegrasse
force-pushed
the
fix/rt1050-hyperflash
branch
from
88bbf06 to
7f23e09
Compare
| L1CACHE_DisableICache(); | ||
| L1CACHE_DisableDCache(); |
There was a problem hiding this comment.
Shouldn't you use the sys_cache_ functions instead? In case the CACHE_MANAGEMENT is turned off completely?
There was a problem hiding this comment.
Yeah, probably best to get this conversion done. Changed to use sys_cache* functions in latest push
danieldegrasse
changed the title
danieldegrasse
force-pushed
the
fix/rt1050-hyperflash
branch
from
7f23e09 to
d3bcc61
Compare
|
NXP specific, re-assigned to @mmahadevan108. |
|
@mmahadevan108 can you take a look? |
Don't access device pointer from critical sections when programming the hyperflash, as this could cause a RWW hazard Signed-off-by: Daniel DeGrasse <ddegrasse@tenstorrent.com>
Disable the cache during erase and programming operations, as cache pre-fetch operations can cause flash access outside of the application's control Also, reduce the SCLK frequency used after erase operations to 200MHz. Without this, the RT1050 appears to hang after flash program operations Signed-off-by: Daniel DeGrasse <ddegrasse@tenstorrent.com>
|
@mmahadevan108 PTAL- have rebased to get a new CI run. |
danieldegrasse
force-pushed
the
fix/rt1050-hyperflash
branch
from
d3bcc61 to
c3971f0
Compare
|
|
Did you guys notice that a similiar fix on another PR I submitted a week before? |
I usually go searching for PRs if I'm about to start fixing a driver- that being said, I think this PR sort of went unnoticed (partially because I haven't been pushing to get it merged) |
Apply a few fixes to the NXP FlexSPI hyperflash driver to prevent flash read access from the core while programming the flash. These fixes were tested with #85254 on an RT1050-EVK, as this is where I encountered the crashes.
Note that the ICACHE/DCACHE disable step is based on what is done here: https://github.com/nxp-mcuxpresso/mcux-sdk-examples/blob/52b428258efda7d5bd8a2ace2195ca828356743a/evkbimxrt1050/driver_examples/flexspi/hyper_flash/polling_transfer/flexspi_hyper_flash_ops.c#L279