-
Notifications
You must be signed in to change notification settings - Fork 20.9k
linux/arm: panic: fatal error: malloc deadlock #1962
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
Did this happen only once? Can you reproduce it? |
Also, which commit is this exactly? I'm guessing it's e165c2d, but would be nice to be certain. |
Got the binary about 2 hours ago.
I have restarted the client and it is still running. Let you know when it panics again. |
Just a random note here, I have been running an Android build a while ago and noticed a silent crash (i.e. no output, process just stopped being) around the same block height (I remember hitting 90K before dying), and @anthony-cros also reported that running on the Tizen platform, geth failed at 90K blocks. All of these platforms seem to be some variation of ARM code. |
@CJentzsch @karalabe please set |
same here, raspberry pi, archlinux arm and latest binary (downloaded yesterday) panic log: https://gist.github.com/donSchoe/12f008f036433d5590ff
and can not run it anymore:
Having the same version as @CJentzsch:
|
Was able to catch the core dump.
|
Same problem |
@donSchoe Please open a new issue with your report. This issue tracks an upstream Go bug. Your crash is related to our implementation. After you open the new issue, please mention @fjl in it too so that he's aware of the trie errors. @SebastianBoyd Please open a new bug report too. As there are two separate issues being reported here, I'm not sure which one of them occurred to you. Open a new one and we'll link it together with a previous issue if it's the case. |
@karalabe I can confirm that I got the same problems 3 times already on an arm binary (created via xgo), geth silently stopped at various blocks (~90K, ~120K, ~150K). I was using the latest release. Any idea what could cause it? Need me to try the GOTRACEBACK procedure as well? |
@anthony-cros Yes please. It's kind of shooting in the dark as we have no idea where to look, but an extra data point cannot hurt. So if you can repro it, all information would be welcome. |
I run FreeBSD arm for go-ethereum and I have not experienced this issue. |
@karalabe OK so I don't have enough privileges to change the ulimit value (no root access on tizen devices), but I exported GOTRACEBACK to crash and resumed geth. It's continuing from where it had stopped and so far it's working but I expect it to die after a few 10K blocks or so. I'll report back ASAP! @tymat , that's interesting, did you also use XGO to create your binary? |
@anthony-cros No. I had to build natively from within the Pi itself using go1.5 FreeBSD arm. Are the Linux builds for ARM made with a cross compiler? |
Yes, @tymat. With xgo. |
Sadly I wasn't able to get any trace :( The good news though is that restarting |
@bobsummerwill Possibly an issue with the cross compiler itself. What is the CPU target? |
@anthony-cros? Just ARM unknown, right? |
Yes @bobsummerwill , this is what we default to |
Hey all, this bug was supposedly fixed upstream in the Go repository. It is scheduled for release in Go 1.5.2, though I'm not sure when that supposed to happen. Until then however I've updated the xgo cross compilation images to support Go development branches, and we've pushed Makefile updates to the geth repo to support selective cross compiles and Go runtime selection. If you wish to try out building with the patched Go runtime, you can do so via:
|
Great! Thanks for the update @karalabe. @anthony-cros Please could you use the 1.5-develop branch while putting together our go-ethereum-cross Dockerfile for doublethinkco/cpp-ethereum-cross#29? |
Sounds good @bobsummerwill , and great job @karalabe , I'm curious to know what the problem was (if it's easy to summarize)? |
Yes - I would be interested too. Is there specific go ticket which was the root cause, @karalabe? Got a URL to share? Thanks! |
it's the fact that it was silent that intrigues me the most |
@bobsummerwill The issue was golang/go#13143. |
Thanks, @fjl :-) |
@CJentzsch Go 1.5.2 is out and contains a fix for this issue. |
cherry pick pbss patches from go-ethereum
using the cross compiled binary (https://build.ethdev.com/builds/ARM%20Go%20develop%20branch/geth-ARM-latest.tar.bz2)
I got a panic: https://gist.github.com/CJentzsch/31aab5b31fae6805f893
The text was updated successfully, but these errors were encountered: