qed: put cond_resched() in qed_dmae_operation_wait()

It is OK to sleep in qed_dmae_operation_wait, because it is called only
in process context, while holding p_hwfn->dmae_info.mutex from one of
the qed_dmae_{host,grc}2{host,grc} functions.
The udelay(DMAE_MIN_WAIT_TIME=2) in the function is too short to replace
with usleep_range, but at least it's a suitable point for checking if we
should give up the CPU with cond_resched().

This lowers the latency caused by 'ethtool -d' from 10 ms to less than
2 ms on my test system with voluntary preemption.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Link: https://patch.msgid.link/20240930201307.330692-5-mschmidt@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

diff --git a/drivers/net/ethernet/qlogic/qed/qed_hw.c b/drivers/net/ethernet/qlogic/qed/qed_hw.c
index 6263f847b6b9265a55681e081655e171073e9170..9e5f0dbc8a07654fe1449c08e7eecade298770d9 100644 (file)
--- a/drivers/net/ethernet/qlogic/qed/qed_hw.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_hw.c
@@ -596,6 +596,7 @@ static int qed_dmae_operation_wait(struct qed_hwfn *p_hwfn)
        barrier();
        while (*p_hwfn->dmae_info.p_completion_word != DMAE_COMPLETION_VAL) {
                udelay(DMAE_MIN_WAIT_TIME);
+               cond_resched();
                if (++wait_cnt > wait_cnt_limit) {
                        DP_NOTICE(p_hwfn->cdev,
                                  "Timed-out waiting for operation to complete. Completion word is 0x%08x expected 0x%08x.\n",