19:30 <xdarklight> webczat: nice, that "fill data with 0xff then erase" test seems like a good confirmation that it works

19:30 <xdarklight> webczat: did you also run Linux for a few days with discard enabled (on the eMMC / SD card partitions)?

19:31 <xdarklight> if you don't get any corruption from that (assuming that you have a working eMMC frequency) then you should send a patch upstream and get feedback from the MMC and Amlogic developers/hackers/interested people

19:34 <webczat> xdarklight: the test was done on u-boot. i imply that it works because: a. mmc erase works in general like core supports it well. b. neither kernel nor uboot have erase code in the driver, so there is nothing erase related in the driver itself to test. c. both drivers work in general and erase is probably not the most complicated/special command that needs special support. and d. tested and works on

19:34 <webczat> uboot without adding any additional code.

19:35 <webczat> so by implication I would assume it works on linux. although yes i could do what you said, but some corruptions could hide themselves

19:36 <xdarklight> webczat: there's always the chance of breaking something (even not touching existing code at all could mean that something breaks because of bad weather...) :)

19:37 <xdarklight> webczat: if there's some hidden corruption which is hard to notice: bad things happen. but if there's corruption that's easy to spot/reproduce then it would be better NOT to enable erase support until the cause is found

19:43 <webczat> xdarklight: okay. i would rather not do this to my still not installed system, but i have no objections and i probably will, especially that I am quite confident that it is tested and works, so have no reason to be afraid (until it breaks hahhaha)

19:44 <xdarklight> :D

19:47 <webczat> well it is just quite interesting for me why no one did that already.

19:52 <lvrp16> flash storage in general is a software minefield because of the dozens of controllers in use

19:55 <lvrp16> at least you have some control with nand, the moment a controller gets slapped on in asia, the battle is lost

20:03 <webczat> lvrp16: but this is testing just one driver. :) they are rather not so different to behave differently in this case?

21:59 <webczat> xdarklight: i have made a patch that enables erase, and i will send it after whatever i wrote in it's description becomes true, that is after I really test linux kernel with discard option