Thread (63 messages) 63 messages, 7 authors, 2025-09-09

Re: [PATCH v6 5/6] tracing: Show inode and device major:minor in deferred user space stacktrace

From: Steven Rostedt <rostedt@goodmis.org>
Date: 2025-08-29 16:49:08
Also in: bpf, lkml

On Fri, 29 Aug 2025 09:28:41 -0700
Linus Torvalds [off-list ref] wrote:

Don't try to "reuse" hashes. Just treat them as opaque numbers.
What do I use to make the hash?

One thing this is trying to do is not have to look up the path name for
every line of a stack trace.

I could just have every instance do the full look up, make a hash from the
path name and build id, and pass the hash to the caller. My worry is the
time it takes to generate that.

Perhaps I could have a hash that maps the pid with the vma->vm_start, and
if that's unique, get the path and build-id and create the hash for that
and send it back to the user. Save the hash for that mapping in the
rhashtable with the pid/vm_start as the key.

Then the code that adds the vma, will see if the pid/vma->start exists, if
it does, return the hash associated with that, if it does not, add it and
trigger the event that a new address has been created.

-- Steve
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help