I've cooked up some patches that maps David's new discard requests to
erase operations on MMC and SD cards. I'm not entirely sure these are
something to keep though as I've been unable to see any performance
increase in keeping blocks erased. Do we have any other reason to keep
it?
During development and test I noticed one issue though; the discard
requests are chopped up into smaller pieces. As the erase commands have
a granularity, sometimes this can mean that parts of the flash are
missed because the discard request isn't split in alignment with the
erase boundaries.
For this reason, I propose that discard requests do not respect the
regular queue restrictions. Those are generally for expressing
limitations in the devices' DMA engines. Since the discard request
carries no data, the DMA engine will never get involved.
--
-- Pierre Ossman
Linux kernel, MMC maintainer http://www.kernel.org
rdesktop, core developer http://www.rdesktop.org
WARNING: This correspondence is being monitored by the
Swedish government. Make sure your server uses encryption
for SMTP traffic and consider using PGP for end-to-end
encryption.
Attachment:
signature.asc
Description: PGP signature
