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
NULL pointer deref of vt->backend in vbp_poke() #2278
Comments
@joshuaspence, could you please install debuginfo packages? |
Sorry, what do you mean by debuginfo packages? |
@joshuaspence debuginfo packages contain symbol information useful for debugging in separate files. But I just realized that the https://packagecloud.io/varnishcache/varnish5 does not contain them, sorry. |
Yep, PackageCloud. |
@joshuaspence could please tell me exactly which package you are using? |
inspecting the binaries from https://packagecloud.io/varnishcache/varnish5/packages/el/7/varnish-5.1.1-1.el7.x86_64.rpm/download with gdb and assuming that the first function called is pool_thread, I get the following estimate of the functions we on the stack (function offset in decimal as returned by gdb disassemble).
|
@nigoroll, see my latest comment in #2269, I don't think installing the debuginfo sub-package will help the back-trace. I'm busy and can't work on the packaging but suggested a fix for future RPM packages. This however, looks like a Debian system. @joshuaspence could you install de38712 from source? This way if the segfault occurs again the back-trace should look similar, but with symbols. |
OK, for https://packagecloud.io/varnishcache/varnish5/packages/debian/jessie/varnish_5.1.1-1_amd64.deb/download things seem to match with no offset for pool_thread
so the annotated stack trace is then:
which matches the segmentation fault if so I think we're here:
so it looks like we're running into
-> it all works out Can't continue with this now, anyone feel free to take it, otherwise I might get back later |
credit for offsetof macro |
@nigoroll it looks like I made the assumption that Thoughts? I can write a patch this evening. |
Thanks for the great work here :) |
So, yesterday evening wasn't quiet enough to have a look so I had a quick one now. This may happen when the backend is deleted, not going cold. I will send a patch today, I expect (again) quiet time ahead. |
Proxy support for backend probes is not in 4.1, so no backport is necessary. |
Originally reported on #2263 by @joshuaspence:
The text was updated successfully, but these errors were encountered: