Message ID | 67985e39f53b56385d79a4f1264cf7f9cacedb58.1742308248.git.robin.murphy@arm.com (mailing list archive) |
---|---|
State | New |
Headers | show |
Series | [Resend] perf/arm-cmn: Remove CMN-600 DTC domain special case | expand |
Hi Robin, On Tue, Mar 18, 2025 at 02:32:10PM +0000, Robin Murphy wrote: > The special case for trying to infer the DTC domain for DTC-adjacent > nodes on CMN-600 is fragile and buggy - currently resulting in subtly > messed up DTC counter allocation - and the theoretical benefit it > offers to a tiny minority of use-cases arguably doesn't outweigh the > inconsistency it offers to others anyway. Just get rid of it. > > Fixes: ab33c66fd8f1 ("perf/arm-cmn: Enable per-DTC counter allocation") > Signed-off-by: Robin Murphy <robin.murphy@arm.com> > --- > drivers/perf/arm-cmn.c | 7 ------- > 1 file changed, 7 deletions(-) You've got "Resend" in the subject, but I can't seem to find the original posting :/ Please can you point me at it? I just want to make sure that I'm not missing patches, as you're unlikely to be the only person affected if that is the case. Cheers, Will
On 18/03/2025 4:17 pm, Will Deacon wrote: > Hi Robin, > > On Tue, Mar 18, 2025 at 02:32:10PM +0000, Robin Murphy wrote: >> The special case for trying to infer the DTC domain for DTC-adjacent >> nodes on CMN-600 is fragile and buggy - currently resulting in subtly >> messed up DTC counter allocation - and the theoretical benefit it >> offers to a tiny minority of use-cases arguably doesn't outweigh the >> inconsistency it offers to others anyway. Just get rid of it. >> >> Fixes: ab33c66fd8f1 ("perf/arm-cmn: Enable per-DTC counter allocation") >> Signed-off-by: Robin Murphy <robin.murphy@arm.com> >> --- >> drivers/perf/arm-cmn.c | 7 ------- >> 1 file changed, 7 deletions(-) > > You've got "Resend" in the subject, but I can't seem to find the original > posting :/ > > Please can you point me at it? I just want to make sure that I'm not > missing patches, as you're unlikely to be the only person affected if > that is the case. Sorry, forgot to explain below the cut - this is a resend from my end because I fat-fingered my LAKML alias the first time around, so I wasn't entirely sure whether you'd get it twice. I guess the error I got was actually from the SMTP server before it delivered anything, phew! Thanks, Robin.
diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c index ef959e66db7c..582f0c0e1783 100644 --- a/drivers/perf/arm-cmn.c +++ b/drivers/perf/arm-cmn.c @@ -2166,13 +2166,6 @@ static int arm_cmn_init_dtcs(struct arm_cmn *cmn) cmn->xps = arm_cmn_node(cmn, CMN_TYPE_XP); - if (cmn->part == PART_CMN600 && cmn->num_dtcs > 1) { - /* We do at least know that a DTC's XP must be in that DTC's domain */ - dn = arm_cmn_node(cmn, CMN_TYPE_DTC); - for (int i = 0; i < cmn->num_dtcs; i++) - arm_cmn_node_to_xp(cmn, dn + i)->dtc = i; - } - for (dn = cmn->dns; dn->type; dn++) { if (dn->type == CMN_TYPE_XP) continue;
The special case for trying to infer the DTC domain for DTC-adjacent nodes on CMN-600 is fragile and buggy - currently resulting in subtly messed up DTC counter allocation - and the theoretical benefit it offers to a tiny minority of use-cases arguably doesn't outweigh the inconsistency it offers to others anyway. Just get rid of it. Fixes: ab33c66fd8f1 ("perf/arm-cmn: Enable per-DTC counter allocation") Signed-off-by: Robin Murphy <robin.murphy@arm.com> --- drivers/perf/arm-cmn.c | 7 ------- 1 file changed, 7 deletions(-)