Quoting Rasmus Villemoes (2021-03-04 15:11:47) > On 04/03/2021 20.15, Stephen Boyd wrote: > > Quoting Matthew Wilcox (2021-03-04 09:00:52) > >> On Mon, Mar 01, 2021 at 09:47:47AM -0800, Stephen Boyd wrote: > >>> Example: > >>> > >>> WARNING: CPU: 4 PID: 3255 at drivers/misc/lkdtm/bugs.c:83 lkdtm_WARNING+0x28/0x30 [lkdtm] (ed5019fdf5e53be37cb1ba7899292d7e143b259e) > >> > >> Would the first 12 characters instead of all 40 make it more palatable > >> without reducing its utility? > > > > I can't seem to request debuginfo from debuginfod without the full 40 > > characters. It's not a git sha1 hash. > > > >> And I feel it should be within the [], so maybe this: > >> > >> WARNING: CPU: 4 PID: 3255 at drivers/misc/lkdtm/bugs.c:83 lkdtm_WARNING+0x28/0x30 [lkdtm ed5019fdf5e5] > >> > > > > Sure I could put the hex numbers inside the brackets. I suspect changing > > %pS or updating the "Modules linked in:" line isn't going to be > > palatable. I've decided to introduce another printk format %pT to print > > the stacktrace > > Can you avoid claiming a new "top-level" %p modifier? Isn't it better to > add a new flag to '%pS', say '%pSb' to include build-id? > I see that %pSR is used in alpha for the stacktrace. I guess we can have %pSb and %pSr then.
