Re: [PATCH 1/2] Unify the handling of stack traces in the consumer
Kris Van Hees <[email protected]>
| Newsgroups | dev.linux.lists.dtrace |
|---|---|
| Message-ID | <[email protected]> |
On Sun, Sep 28, 2025 at 07:52:52PM -0400, Eugene Loh wrote:
> On 9/28/25 19:26, Kris Van Hees wrote:
>
> > On Sun, Sep 28, 2025 at 01:50:02AM -0400, Eugene Loh wrote:
> >
> > > For kernel stacks, I guess size is always 8. Could mention that? Also not a
> > > big deal.
> > I am not sure what you mean by this? The size of the data that is set aide to
> > store the stack trace depends on the number of frames that is requested.
>
> Right. I mean the size of a PC.
>
> The code used to have a function
>
> int
> dt_print_stack(dtp, fp, format, addr, depth, size)
> {
> Â Â Â Â Â Â Â for (i = 0; i < depth; i++) {
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â switch (size) {
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â case sizeof(uint32_t):
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â /* LINTED - alignment */
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pc = *((uint32_t *)addr);
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â case sizeof(uint64_t):
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â /* LINTED - alignment */
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pc = *((uint64_t *)addr);
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â default:
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â return dt_set_errno(dtp, EDT_BADSTACKPC);
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â }
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â addr += size;
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â [...]
> Â Â Â Â Â Â Â }
> }
>
> (Lots of code redacted.) Anyhow, we feed in a "size" argument. The ported
> function is
>
> int
> dt_print_stack_kernel(dtp, fp, format, addr, indent, depth)
> {
> Â Â Â Â Â Â Â for (i = 0; i < depth; i++) {
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pc = *((uint64_t *)addr);
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â if (pc == 0)
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â addr += sizeof(pc);
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â [...]
> Â Â Â Â Â Â Â }
> }
>
> which assumes 8-byte values. Again, no big deal.
Oh, that.... Well, since we no longer support 32-bit anyway, it seems hardly
relevant to keep code around that handles 32-bit PCs. If we were to bring
back tracing of 32-bit apps on a 64-bit system, we can re-introduce code for
it, hopefully in a more clean manner, but that will require additional code in
various other places I believe.