| Message ID | 20221117104957.254648-1-bmasney@redhat.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | qcom: add basic interconnect support to UFS | expand |
On 17/11/2022 11:49, Brian Masney wrote: > This patch set adds very basic support for the interconnect framework > to the Qualcomm portion of the UFS framework since the firmware on > these platforms expects the interconnect votes to be present. The > maximum throughput is requested to match what's already done in a few > other drivers. > > Here's the relevant entries from the interconnect_summary file in > debugfs that shows the two ICC paths are setup for the first UFS > host controller on the SA8540p automotive board (sc8280xp). I wonder whether this is solving the same or orthogonal problem as my old patchset here: https://lore.kernel.org/all/20220513061347.46480-1-krzysztof.kozlowski@linaro.org/ Best regards, Krzysztof
On 17/11/2022 13:12, Krzysztof Kozlowski wrote: > On 17/11/2022 11:49, Brian Masney wrote: >> This patch set adds very basic support for the interconnect framework >> to the Qualcomm portion of the UFS framework since the firmware on >> these platforms expects the interconnect votes to be present. The >> maximum throughput is requested to match what's already done in a few >> other drivers. >> >> Here's the relevant entries from the interconnect_summary file in >> debugfs that shows the two ICC paths are setup for the first UFS >> host controller on the SA8540p automotive board (sc8280xp). > > I wonder whether this is solving the same or orthogonal problem as my > old patchset here: > > https://lore.kernel.org/all/20220513061347.46480-1-krzysztof.kozlowski@linaro.org/ More or less it does. Vendor kernel scales both paths according to the gear selected. I was surprised to see just two entries there. sdm845 has 22 entries in its msm-bus scaling table. What was the reason for just two entries in your case? What was the net result for that patchset? Is it going to be merged anytime? I think we can start with just a version of this patchset that enables static ICC config and then upgrade that with proper OPP tables, WDYT? (I wrote 'a version' since I had to modify the patch to set avg_bw instead of setting the peak_bw and to pass different values instead of UINT_MAX, I'll send it).