On Fri, Mar 20, 2026 at 10:43 AM Louis Chauvet <[email protected]> wrote: > > On Fri, 13 Mar 2026 07:20:29 -0600, Jim Cromie <[email protected]> wrote: > > [...] > > mismatch in intel_pipe_config_compare(), the resulting UART storm > > causes a hard timeout after 20 minutes (see below the snip). > > > > To fix this, change __drm_printfn_dbg() to use the explicit > > drm_debug_enabled_instrumented() instead. This ensures the bit-test is > > performed at runtime for this unguarded helper, stopping the UART storm. > > Why don't use use drm_debug_enabled directly? Does it make sense to call > the _instrumented when [2] is not enabled?
my reason is so that we can turn on the pr-debug, and see how often this happens. If it happens a lot, we can look for ways to replace the bit-test with a static-key guard. That said, we could use drm_debug_enabled(), we'd lose the counting ability when DRM_USE_DYNAMIC_DEBUG=n, but still have it for =y. So Im happy with either way. > > > > > -- > Louis Chauvet <[email protected]>
