On 18.01.2019 09:50, Martin Husemann wrote: > Sorry, I completely fail to parse this - can you start from scratch and > just describe the problem you think you are seeing? > > Martin > If we trigger SIGBUS with the trigger_bus() function (mentioned in the initial message), the core(5) dumping process is interrupted. It's interrupted because we cannot access the pages that are not assigned (which is the cause of the original crash emitting SIGBUS). Stacktrace in the original patch shows that we receive EINVAL in coredumping routines. I don't know if there is a clean way to handle it so I've added a hack to handle this failure with handing the EINVAL value returned from genfs_getpages(). It's a hack but works. I'm open to a better solution if there is any.
Attachment:
signature.asc
Description: OpenPGP digital signature