On Mon, 5 May 2025 at 13:55, Jann Horn <jannh@xxxxxxxxxx> wrote: > > On Mon, May 5, 2025 at 1:14 PM Christian Brauner <brauner@xxxxxxxxxx> wrote: > > The new coredump socket will allow userspace to not have to rely on > > usermode helpers for processing coredumps and provides a safer way to > > handle them instead of relying on super privileged coredumping helpers. > > > > This will also be significantly more lightweight since no fork()+exec() > > for the usermodehelper is required for each crashing process. The > > coredump server in userspace can just keep a worker pool. > > I mean, if coredumping is a performance bottleneck, something is > probably seriously wrong with the system... I don't think we need to > optimize for execution speed in this area. It can be a bottleneck because it blocks restarting processes until it completes, which directly impacts SLAs if the crashed process is relevant for those purposes, and it often is, we have several use cases where that's the case. The usermode helper costs hundreds of ms in latency, so any optimization in that area is very welcome and worth mentioning IMHO.
