]> git.baikalelectronics.ru Git - kernel.git/commit
igb: Make DMA faster when CPU is active on the PCIe link
authorKai-Heng Feng <kai.heng.feng@canonical.com>
Tue, 21 Jun 2022 22:10:56 +0000 (15:10 -0700)
committerJakub Kicinski <kuba@kernel.org>
Thu, 23 Jun 2022 01:46:24 +0000 (18:46 -0700)
commit157fc3e6bedf1113da0b52e2f2c0004c48174c2f
treee8cf0874494f40f46dc6f5e5bccbe9853e4bedc8
parent6c9acb91bc7459fae6a9b901dfba066470801f0a
igb: Make DMA faster when CPU is active on the PCIe link

Intel I210 on some Intel Alder Lake platforms can only achieve ~750Mbps
Tx speed via iperf. The RR2DCDELAY shows around 0x2xxx DMA delay, which
will be significantly lower when 1) ASPM is disabled or 2) SoC package
c-state stays above PC3. When the RR2DCDELAY is around 0x1xxx the Tx
speed can reach to ~950Mbps.

According to the I210 datasheet "8.26.1 PCIe Misc. Register - PCIEMISC",
"DMA Idle Indication" doesn't seem to tie to DMA coalesce anymore, so
set it to 1b for "DMA is considered idle when there is no Rx or Tx AND
when there are no TLPs indicating that CPU is active detected on the
PCIe link (such as the host executes CSR or Configuration register read
or write operation)" and performing Tx should also fall under "active
CPU on PCIe link" case.

In addition to that, commit 979bf8237421 ("igb: Move DMA Coalescing init
code to separate function.") seems to wrongly changed from enabling
E1000_PCIEMISC_LX_DECISION to disabling it, also fix that.

Fixes: 979bf8237421 ("igb: Move DMA Coalescing init code to separate function.")
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20220621221056.604304-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
drivers/net/ethernet/intel/igb/igb_main.c