Re: [PATCH] drm/msm: handle for EPROBE_DEFER for of_icc_get
On Wed, Jul 01, 2020 at 01:13:34PM -0400, Jonathan Marek wrote: > On 7/1/20 1:12 PM, Matthias Kaehlcke wrote: > > Hi Jonathan, > > > > On Tue, Jun 30, 2020 at 11:08:41PM -0400, Jonathan Marek wrote: > > > Check for EPROBE_DEFER instead of silently not using icc if the msm driver > > > probes before the interconnect driver. > > > > Agreed with supporting deferred ICC probing. > > > > > Only check for EPROBE_DEFER because of_icc_get can return other errors > > > that > > > we want to ignore (ENODATA). > > > > What would be the -ENODATA case? > > > > The of_icc_get for the ocmem_icc_path can return -ENODATA when the ocmem > path is not specified (it is optional and only relevant for a3xx/a4xx). Thanks for the clarification! In this case it seems reasonable to me to return any error for the 'gfx-mem' path and all errors except -ENODATA for 'ocmem'.
Re: [PATCH] drm/msm: handle for EPROBE_DEFER for of_icc_get
On 7/1/20 1:12 PM, Matthias Kaehlcke wrote: Hi Jonathan, On Tue, Jun 30, 2020 at 11:08:41PM -0400, Jonathan Marek wrote: Check for EPROBE_DEFER instead of silently not using icc if the msm driver probes before the interconnect driver. Agreed with supporting deferred ICC probing. Only check for EPROBE_DEFER because of_icc_get can return other errors that we want to ignore (ENODATA). What would be the -ENODATA case? The of_icc_get for the ocmem_icc_path can return -ENODATA when the ocmem path is not specified (it is optional and only relevant for a3xx/a4xx). If the 'interconnects' property is not specified of_icc_get() returns NULL, shouldn't all (or most) errors be propagated rather than staying silent? Thanks Matthias
Re: [PATCH] drm/msm: handle for EPROBE_DEFER for of_icc_get
Hi Jonathan, On Tue, Jun 30, 2020 at 11:08:41PM -0400, Jonathan Marek wrote: > Check for EPROBE_DEFER instead of silently not using icc if the msm driver > probes before the interconnect driver. Agreed with supporting deferred ICC probing. > Only check for EPROBE_DEFER because of_icc_get can return other errors that > we want to ignore (ENODATA). What would be the -ENODATA case? If the 'interconnects' property is not specified of_icc_get() returns NULL, shouldn't all (or most) errors be propagated rather than staying silent? Thanks Matthias
