IRC logs for #openrisc Friday, 2014-06-13

--- Log opened Fri Jun 13 00:00:32 2014
stekernblueCmd: https://github.com/openrisc/or1k-gcc/blob/or1k/gcc/config/or1k/or1k.c#L132 <- any insight on how to get rid of that?04:31
stekernit clutters up all functions with a redundant load+store04:32
stekernmy smp wip crashes elegantly when in 3.15...04:44
_franck_stekern: ok I will05:52
stekern_franck_: how did it 'crash after a while'?06:01
olofkstekern: If you're using emacs you can go to the beginning of the line and press Ctrl+k to get rid of it06:02
stekernolofk: actually, I think we need to get rid of all of this: https://github.com/openrisc/or1k-gcc/blob/or1k/gcc/config/or1k/or1k.c#L132-L13606:28
stekern(I just found out you can do ranges like that :)06:28
olofkstekern: C-space, C-n, C-n, C-n, C-n, C-w08:30
olofkOr just M-x tetris and forget about the whole thing08:34
stekernwhy isn't python's partition always returning 'pre-sep' 'sep' 'post-sep'08:45
stekern?08:45
stekernbut instead if sep: 'pre-sep' 'sep' 'post-sep' else: 'post-sep' '' ''08:46
olofkIs that lpartition or rpartition? There's two, right?08:46
stekern...well, I guess it depends how you look at it, maybe they though that my post-sep is pre-sep if there is no sep08:46
stekernthen it's actually logical08:47
stekern..as usual, wenting things on IRC helps ;)08:47
stekernfor my particular use-case, it sucks that it works like that though08:47
olofkI'm working on my first awk(ward) script. Weird shit08:47
olofkbut if not sep: pre-sep == post-sep?08:48
stekern"comma,seperated".partition(',') => 'comma' ',' 'seperated'08:50
stekern"comma.seperated".partition(',') => 'comma.seperated' '' ''08:51
olofkThat seems logical to me08:56
stekernyes, it is08:56
olofkImported the trace log to a sqlite db now. Time to do some post processing analysis on that :)08:58
olofkoh... can awk read a hex string and convert it to decimal btw?09:03
stekernwhat are you doing?09:03
stekernsqlite dbs... sounds scary09:03
olofkGetting some statistics on where we spend cycles09:09
olofkand trying to find out why my simulation just stops09:09
stekernusing sqlite as a middle hand still sounds scary ;)09:09
olofkNever used it before. Let's see how that turns out :)09:11
olofkgrepping through 2.5GB plain text logs is a bit scary as well :)09:12
stekerngrep is fast09:12
stekern;)09:12
stekernare you still running in icarus?09:16
stekernI've found the reason why my smp boot crashes, not sure how to fix it yet though09:18
stekernthe wrong r10 get restored when returning from fork09:18
stekernso suddenly both precessors are running on the same kernel stack09:19
stekern...the failures are spectacular and extremely subtle09:19
stekerns/precessors/processors09:20
stekernbah... it was an old bug I already fixed, but I missed to merge that thing into 3.1510:18
olofkwrong r10? How many r10 do you have?10:31
stekernwell, one for each thread10:36
stekern;)10:36
stekernit might occasionally be saved on the stack though10:36
stekern+ one for userspace10:36
olofkOh god. I realize that I haven't missed sql at all since I used it last11:32
olofkWhy does GCC push r1 to the tack when it doesn't use it in a function?12:20
stekernbecause I just played tetris...12:23
olofk:)12:23
olofkOr is r1 always saved? I thought I found some functions that didn't save r112:28
stekernyes, the answer is in: https://github.com/openrisc/or1k-gcc/blob/or1k/gcc/config/or1k/or1k.c#L132-L13612:30
stekernit's always saved when r2 isn't used as a framepointer12:31
olofkahhh... I could have sworn that I've seen that URL before :)12:44
stekernhence the tetris joke ;)13:02
olofkHaha, I'm slow :)13:27
_franck__cool, I have a CFI NOR peripheral in or1ksim (at least it works with barebox)13:43
_franck__stekern: I tested your kernel during my lunch time. I had a kernel panic during the boot, before busybox started. My be some configuration problem on my side13:47
stekern_franck__: in or1ksim or on a real board?14:35
olofk_franck__: Nice. Can you make a VPI wrapper for it as well? :)14:36
_franck__in or1ksim14:42
_franck__for now I just have CFI detection and READ command14:43
stekernoh... do you have a recent version with l.swa/l.lwa support?14:43
_franck__no I don't14:43
stekernI just realised that the busybox I put there has those14:43
stekernbecause the old busybox won't work, I didn't bring in the deprecated syscalls14:43
_franck__sorry I was thinking about my CFI flash14:44
stekernah ;)14:44
_franck__so I have a recent version of the toolchain with swa/lwa14:44
_franck__and a real boar14:44
_franck__board14:44
stekernbut, I asked about or1ksim, not toolchain ;)14:45
_franck__ok :)14:45
stekernbut if you have a busybox somewhere, you can throw in that instead14:46
stekernI mean, from when you tested the latest uclibc14:46
stekernthe busybox I put there was compiled against musl14:47
_franck__I use the same busybox as I use with the vanilla kernel (the kernel without with rebase)14:47
stekernok...14:47
stekerncan you throw me the vmlinux14:47
_franck__with my vanilla kernel, I got an unaligned access after some time14:48
_franck__with your it didn't start busybox14:48
_franck__I'll give you this tonight14:48
olofkHow is the delay slot-less version handled in Linux? Are there any ifdef's, or is all asm written to handle both cases?14:50
stekernno14:53
olofkCan -nd run Linux at all?15:31
stekernno15:51
_franck_olofk: we are trying to clean patch list for openocd. Would you mind to take a look at this: http://openocd.zylin.com/#/c/838/18:45
_franck_if you don't want to work on it, please "abandon" it18:45
olofk_franck_: I can't figure out my username/password. Can anyone else abandon it for me?18:50
_franck_ok I'll tell Andreas to do it for you18:53
olofkI reset the password, so I'll do it18:56
_franck_ok thanks18:57
_franck_stekern: https://www.dropbox.com/s/yncl1ekn570x118/vmImage19:02
stekerndanke schön19:02
stekernummm19:02
stekernno, nothing19:03
_franck_I got it ;)19:03
stekernI was more referring to, what are you giving me?19:04
stekernit's not an or1k elf at least19:04
_franck_really ? :) let me check19:04
stekernand why is the name vmImage?19:05
_franck_wrong file, that's a uboot one19:05
_franck_is your clock 50MHz ?19:06
_franck_https://www.dropbox.com/s/kdxi4dsg3dnaaon/vmlinux19:07
stekernmerci beaucoup19:07
_franck_de rien19:07
_franck_vmImage is coming from here: https://github.com/fjullien/linux/commit/3a17fed6b7bc5b6c6f9fc365419d0b34e77234a519:07
stekernisn't uImage the more common name?19:08
stekernI mean, the target name most other archs are using?19:09
_franck_I couldn't find what's the rule. There isa _lot_ of names for that19:09
stekern=)19:09
_franck_we choose when it'll be tested and ready to commit. I also wanted to implement zImage (self decrompressed kernel) but I'm sure it worth it19:11
_franck_BTW, when I first compile your kernel with frame buffer and frame buffer console enable I was missing asm/vga.h19:13
_franck_I just added something like this http://lxr.free-electrons.com/source/arch/x86/include/asm/vga.h19:13
stekernhmm, I think that should be fixed already... maybe I missed that patch somehow...19:15
stekernbut... your kernel crashes in mem_serial_out19:15
_franck_why would my uart do that ?19:17
stekernI've got no idea, and another strange artifact is that your kernel prints: "Unpacking initramfs"19:17
_franck_I'll double check my config19:18
stekerncan you throw me your busybox too?19:19
_franck_https://www.dropbox.com/s/4v0c94oqsf21igg/busybox19:20
stekernkiitos paljon19:22
_franck_don't know this one ;)19:22
stekernhttp://pastie.org/928744219:26
_franck_so my busybok is working19:27
stekernyes, but where did you get your kernel from: Linux version 3.4.0-or1ksim-76951-g18a07c3-dirty19:28
stekern?19:28
_franck_ahh, don't tell me I'm working on the wrong kernel !19:29
_franck_stupid me !19:29
_franck_I didn't checkout the good branch after I clone your repo19:30
stekernheh, ok19:30
_franck_I knew that was strange when I make oldconfig, there was a lot of diffs19:30
_franck_I'll do it again now19:31
stekernfwiw, I pushed some lingering patches to the master branch too, ontop of the ones that were already in jonas repo19:31
_franck_ http://pastie.org/9287475 --> not working, brb19:42
stekernwhat are you doing differently from me? =)19:44
stekernbase_baud = 4166666 <- wut?19:45
_franck_I'm running at 66MHz19:56
stekernah, right. that's what that was showing20:01
stekernwanna give me that vmlinux too? =)20:03
_franck_only head.S is different between the working and not working kernel it must be something big in front of my eyes20:09
_franck_but I don't see it20:09
stekernwhat's at c00182ec20:14
_franck_don't go further with my "working" kernel I get  http://pastie.org/928754420:19
_franck_why ?20:19
_franck_enough for tonight, I'm tired20:19
stekernwhat busybox are you using now?20:20
stekernit's not the one you gave me20:20
_franck_true20:20
_franck_at least it starts busybox20:20
stekerninit started: BusyBox v1.22.1 (2014-06-08 22:11:20 CEST) vs init started: BusyBox v1.22.1 (2014-06-10 23:01:17 CEST)20:20
stekernyes, but it looks like you compiled the one you are using now against an old uclibc20:21
stekernand the kernel you are using doesn't have the syscalls that that one is using20:22
_franck_I didn't update my toolchain since then...20:23
_franck_too bad we can't identify which uClibc it is from the toolchain binary20:23
_franck_can we ?20:23
stekernyes you did update your toolchain after 2014-06-08: http://juliusbaxter.net/openrisc-irc/%23openrisc.2014-06-09.log.html#t21:0420:25
_franck_you're better than the NSA20:25
_franck_:)20:25
stekern=)20:26
_franck_with the good initramfs, my "working" kernel works20:30
_franck_not yours20:33
_franck_it has something to do with SPARSE_IRQ or WISHBONE_BUS_BIG_ENDIAN20:42
_franck_I now have your tree booting. I just need to revert changes I did to see which one make things work20:43
_franck_(I have also head.S to revert and some other minor changes)20:44
_franck_https://www.dropbox.com/s/kdxi4dsg3dnaaon/vmlinux <-- it craches "after some time"20:45
stekernis that vanilla or the openrisc-3.15-merge, or my masteR?20:45
_franck_it's openrisc-3.15-merge where I changes some stuff to make it work20:46
stekernok20:47
_franck_this is where am I now: http://pastie.org/928759520:47
_franck_I need to revert those changes one after the other to find what make things crash20:48
stekernI meant openrisc-3.15-rebase, but I bet you just copied my mistake ;)20:48
_franck_yes rebase20:49
stekernI don't know what that SPARSE_IRQ thing does, most of the original commit is upstreamed20:49
stekernso, the remains that's still there might very well be wrong20:50
stekern...but they were still in jonas master tree20:50
_franck_we'll find out tomorrow20:50
_franck_stay tuned20:50
stekernI will20:50
_franck_good night20:51
stekernfor the record, this was the original commit: http://git.openrisc.net/cgit.cgi/jonas/linux/commit/?id=24b9718a6b5ac0b03ce360d1bfe115bb2010e0f720:52
stekernyour vmlinux haven't crashed "after a while" here btw20:53
stekernbut I don't know if I'm holding it right20:53
stekernaha! my smp crashes was due to a wb_sdram_ctrl bug23:02
stekernI wonder why olofk's bench didn't catch that one... it's a write followed by a read23:04
--- Log closed Sat Jun 14 00:00:33 2014

Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!