Granite Rapids Idle/P-State Work For Linux 6.12 & An Important AMD Energy Reporting Fix
([Hardware] 2 Hours Ago
Linux 6.12 Power Management)
- Reference: 0001486650
- News link: https://www.phoronix.com/news/Linux-6.12-Intel-AMD-PM-Early
- Source link:
Queued yesterday into the Linux power management subsystem's "linux-next" branch ahead of the Linux 6.12 kernel are some important additions for Intel processors and also an important fix for AMD processors to enjoy package power monitoring.
New on the Intel side is [1]Intel_Idle driver support for Granite Rapids Xeon being queued. The patch message explains of the Intel Idle CPU handling for upcoming Granite Rapids:
"Add Granite Rapids Xeon C-states, which are C1, C1E, C6, and C6P.
Comparing to previous Xeon Generations (e.g., Emerald Rapids), C6 requests end up only in core C6 state, and no package C-state promotion takes place even if all cores in the package are in core C6.
C6P requests also end up in core C6, but if all cores have requested C6P, the SoC will enter the package C6 state."
It's a bit surprising the Granite Rapids support for the Intel_Idle driver didn't land months ago. Also on the Granite Rapids side is [2]this patch also queued yesterday ahead of Linux 6.12. That patch for the Intel P-State CPU frequency scaling driver gets out-of-band (OOB) mode support working for both Granite Rapids and Sapphire Rapids Xeon Scalable processors. Emerald Rapids OOB support for Intel P-State was already in place but older Sapphire Rapids and newer Granite Rapids was missing the IDs.
Also on the Intel power management side for Linux 6.12 that was queued up yesterday is [3]Arrow Lake U support for the Intel RAPL driver for power monitoring.
Over on the AMD side entering the power management queue yesterday is [4]this fix for AMD energy package event reporting. This is important for proper RAPL power reporting to avoid inaccurate power readings due to a regression since Linux 6.10. (An issue I originally encountered and reported and already worked around for all my recent CPU power reporting.)
Still yet to be queued up for Linux power management's linux-next or as 6.11 fixes is [5]this one line patch for RAPL support on AMD Family 1Ah (Zen 5) processors.
We'll see what more power management work is readied in time for the upcoming Linux 6.12 cycle.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=370406bf5738dade8ac95a2ee95c29299d4ac902
[2] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=3ca2a3d1e7271549985dc57c8106bb07b3134ecb
[3] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=eca0f1b0bbf9f1082993206e3430d9da391be366
[4] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=26096aed255fbac9501718174dbb24c935d8854e
[5] https://lore.kernel.org/linux-pm/20240719101234.50827-1-Dhananjay.Ugwekar@amd.com/
New on the Intel side is [1]Intel_Idle driver support for Granite Rapids Xeon being queued. The patch message explains of the Intel Idle CPU handling for upcoming Granite Rapids:
"Add Granite Rapids Xeon C-states, which are C1, C1E, C6, and C6P.
Comparing to previous Xeon Generations (e.g., Emerald Rapids), C6 requests end up only in core C6 state, and no package C-state promotion takes place even if all cores in the package are in core C6.
C6P requests also end up in core C6, but if all cores have requested C6P, the SoC will enter the package C6 state."
It's a bit surprising the Granite Rapids support for the Intel_Idle driver didn't land months ago. Also on the Granite Rapids side is [2]this patch also queued yesterday ahead of Linux 6.12. That patch for the Intel P-State CPU frequency scaling driver gets out-of-band (OOB) mode support working for both Granite Rapids and Sapphire Rapids Xeon Scalable processors. Emerald Rapids OOB support for Intel P-State was already in place but older Sapphire Rapids and newer Granite Rapids was missing the IDs.
Also on the Intel power management side for Linux 6.12 that was queued up yesterday is [3]Arrow Lake U support for the Intel RAPL driver for power monitoring.
Over on the AMD side entering the power management queue yesterday is [4]this fix for AMD energy package event reporting. This is important for proper RAPL power reporting to avoid inaccurate power readings due to a regression since Linux 6.10. (An issue I originally encountered and reported and already worked around for all my recent CPU power reporting.)
Still yet to be queued up for Linux power management's linux-next or as 6.11 fixes is [5]this one line patch for RAPL support on AMD Family 1Ah (Zen 5) processors.
We'll see what more power management work is readied in time for the upcoming Linux 6.12 cycle.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=370406bf5738dade8ac95a2ee95c29299d4ac902
[2] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=3ca2a3d1e7271549985dc57c8106bb07b3134ecb
[3] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=eca0f1b0bbf9f1082993206e3430d9da391be366
[4] https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=linux-next&id=26096aed255fbac9501718174dbb24c935d8854e
[5] https://lore.kernel.org/linux-pm/20240719101234.50827-1-Dhananjay.Ugwekar@amd.com/
phoronix