IRC log of #cubox of Wed 06 Sep 2017. All times are in CEST < Back to index

09:04 topi`> the UBOOT on my mbin lacks "sata" commands, there's just scsi and nothing else. How do I boot it from SATA?
09:05 topi`> it seems the UBOOT on mbin is fairly specialised, it has commands like "mss_boot"
09:21 jnettlet[m]> topi`: AHCI mode sata is accessed via the SCSI command
09:21 jnettlet[m]> same protocol different connector basically
09:22 jnettlet[m]> Ke: will get you those. Sorry been very focused on some other work.
09:49 topi`> OK
09:53 topi`> hmm, "ext2ls scsi 0:5" says Bad device scsi 0
09:54 topi`> ah! it was scsi device 8
09:54 topi`> seems to work, let's load the zImage
09:55 topi`> Image-4.4.8-mcbin-marvell, what format is this? uImage?
09:55 Ke> does ARM64 now support zImages, I heard there was no decompression implemented in the kernel at linux-4.9
09:55 Ke> what we used on ***board was just Image IIRC, anything else refused to boot
09:57 topi`> yeah, I had to use "booti" command to get it to boot
09:58 topi`> ata3: SATA link up 6.0 Gbps (SStatus 133 SControl 300)
09:58 topi`> seems to work
10:00 Ke> nice
10:01 topi`> can't log in. Maybe I mounted the wrong partition as root
10:01 topi`> yes. THe old password worked.
10:03 agraf> Ke: why would you want the kernel to self-compress?
10:03 suihkulokki> Ke: on arm64 bootloader is expected to handle decompressing. for example grub does it fine.
10:03 topi`> I might want to set up an emergency root on mmcblk0 so that in case of SATA malfunction, I can still boot the board somehow
10:03 Ke> agraf: in our case we did not have documentation and there seemed to be awfully few places to put the kernel without killing the bootloader
10:04 topi`> is there any easy way to reset the board electrically remotely? the board goes into a dark server room
10:04 topi`> maybe I'll set up a raspberry pi next to it, with a fixed ip address
10:05 Ke> agraf: in well documented case bootloader decompression is most of the time just as fine, just as long as bootloader does not occupy the physical memory you would want for kernel
10:06 Ke> in our case the easy solution was to strip kernel of unneeded parts so did not investigate better solutions
10:06 topi`> the brute approach :)
10:06 agraf> Ke: i see, so you wanted to reduce the chance to clash with boot loader reserved memory by having the kernel image compressed
10:06 Ke> agraf: no we literally tried a couple of locations and boot failed
10:07 agraf> Ke: in a sane world you have uefi memory maps which tell you exactly what regions are available - that way you can always take a region that is guaranteed to be unused
10:07 agraf> Ke: what boot loader did you have on that system?
10:07 Ke> u-boot -variant
10:07 Ke> also no uefi
10:07 agraf> Ke: k, i see :)
10:07 agraf> Ke: well - recent u-boot supports uefi interfaces
10:08 agraf> Ke: but that doesn't mean that the board vendor went through the effort to actually mark all used dram ranges as reserved (atf, random embedded controllers, etc)
10:08 Ke> we even had the sources, we could have just recompiled
10:09 Ke> anyway recompiling the kernel was the solution we were most familiar with anyway
10:10 agraf> sure, that doesn't help a lot if you can't load it though ;)
10:10 topi`> stick with the familiar things
10:10 topi`> unless you're willing to spend the whole month on the thing ;)
10:10 agraf> Ke: but yeah, going forward a good bunch of that stuff *should* get easier, let's see how much of it plays out :)
10:11 Ke> hopefully
10:11 agraf> suihkulokki: btw, have you see? rob got the uefi shell working in u-boot :)
10:12 topi`> is uefi property of Intel, like EFI is?
10:13 agraf> nope - that's what the u stands for ;)
10:13 agraf> u means "not proprietary" ;)
10:15 topi`> Unproprietary
11:16 Ke> jnettlet[m]: and thanks, that's fine =o)
11:18 suihkulokki> agraf: cool :)
12:01 topi`> SATA link 1 timeout.
12:02 topi`> I get this twice, and it cannot find my SATA drive which is in sata 8
12:02 topi`> unless I manually say "scsi rescan"
12:02 topi`> any idea?
17:39 jnettlet[m]> topi` not sure. Haven't tested that. Perhaps spin up time.