| luc4 | Hello! I’d need to git bisect the kernel being used in the raspberry pi. I’m trying to follow the manuals, but I’m not sure I’m doing the right thing. I expected a debian/ dir to be in every commit, but it seems this commit, for instance, is not including it: https://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux-raspi/+git/hirsute/tree/?id=84faf29b2d529a10cd71e609ddb237a5af8551a4. Is this the wrong git? Or maybe I should expect the debian/ | 10:49 |
|---|---|---|
| luc4 | not to be there? In that case can you point me to the proper manual to follow? Thanks! | 10:49 |
| juergh | @luc4, https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel | 12:00 |
| juergh | luc4, ^^ | 12:01 |
| juergh | bisecting a derivative is not trivial. especially raspi. | 12:02 |
| === luc4_ is now known as luc4 | ||
| luc4 | juergh: that is the page I’m referring to. The debian/ directory is needed, but I do not see it in many commits. | 12:14 |
| luc4 | juergh: I can build with a regular “make”, but I guess I’m only building vmlinuz. Can I simply replace that in /boot/firmware in rpi? Can I ignore the initrd? | 12:15 |
| juergh | luc4, you can't build an Ubuntu kernel by simply running 'make'. | 12:16 |
| juergh | the raspi kernel is a derivative of the master (hirsute) kernel. we take the master hirsute kernel and but the raspi specific patches and configs on top of it. | 12:17 |
| juergh | you cant simply bisect. if you checkout a regular upstream commit than you don't have all the raspiness and that kernel will not boot at all. | 12:17 |
| luc4 | juergh: I see, so I can’t bisect. Thank, at least I won’t try something that cannot be done :-) | 12:19 |
| juergh | luc4, it can be done but it's painful. why do you want to bisect? did you find a regression? | 12:20 |
| luc4 | juergh: yes, usb issue: https://github.com/raspberrypi/linux/issues/3981 | 12:22 |
| gitbot | raspberrypi issue 3981 in linux " RPi 4B, external HDD randomly freezes" [Open] | 12:22 |
| juergh | oh dear. this has been going on forever. and you're hitting this issue? | 12:27 |
| * ogra is entertained by people randomly adding and removing arm_64bit=1 to config.txt in that issue ... | 12:28 | |
| luc4 | juergh: yes, heavily. Sometimes 3 hours and the hd is gone. | 12:30 |
| juergh | ubuntu kernel? single or dual HDD? HDD or SSD? | 12:30 |
| luc4 | juergh: I guess both ubuntu and raspberry os. I’m using ubuntu. What do you mean by single? HDD, not SSD. | 12:31 |
| * ogra would first try to add "usbcore.autosuspend=-1" to the kernel cmdline ... if there isnt traffic on the USB for a while when the system idles, the bus might go into suspend | 12:31 | |
| juergh | some people have 2 HDDs attached. | 12:32 |
| luc4 | juergh: I have 1. But I can plug up to 6, which is how I planned to bisect. | 12:32 |
| luc4 | juergh: otherwise I may have to wait for days to reproduce. | 12:33 |
| juergh | luc4, what the HDD IO workload? | 12:34 |
| luc4 | juergh: I could reproduce even on unmounted HDDs. | 12:34 |
| luc4 | juergh: does not seem to be related to workload. | 12:34 |
| juergh | luc4, which Ubuntu kernels show this? | 12:36 |
| luc4 | juergh: if not possible to bisect on Ubuntu nevermind, I’ll have to somehow switch to raspberry os, hoping it has a 64 bit version. Otherwise I won’t bisect :-) | 12:36 |
| luc4 | juergh: I’m pretty sure 5.4 is not affected, 5.8 is. | 12:36 |
| juergh | luc4, you will have similar bisect issues with the raspi os kernel. | 12:38 |
| juergh | I'll see if I can reproduce this. | 12:38 |
| luc4 | juergh: I’ll try. If impossible, I’ll give up. | 12:39 |
| luc4 | juergh: thanks! | 12:39 |
| juergh | luc4, and your HDD is externally powered I assume? | 12:39 |
| luc4 | juergh: yes, it is. | 12:40 |
| luc4 | ogra, juergh: I missed the advice. I already tried to disable autosuspend in proc. I could still reproduce. | 12:46 |
| ogra | k. was just a thought | 12:46 |
| === cpaelzer_ is now known as cpaelzer | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!