Re: [PATCH V2 1/4] drivers: qcom: rpmh-rsc: simplify TCS locking

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Jul 23 2019 at 12:22 -0600, Stephen Boyd wrote:
Quoting Lina Iyer (2019-07-22 14:53:37)
From: "Raju P.L.S.S.S.N" <rplsssn@xxxxxxxxxxxxxx>

The tcs->lock was introduced to serialize access with in TCS group. But,
drv->lock is still needed to synchronize core aspects of the
communication. This puts the drv->lock in the critical and high latency
path of sending a request. drv->lock provides the all necessary
synchronization. So remove locking around TCS group and simply use the
drv->lock instead.

This doesn't talk about removing the irq saving and restoring though.
You mean for drv->lock? It was not an _irqsave/_irqrestore anyways and
we were only removing the tcs->lock.

Can you keep irq saving and restoring in this patch and then remove that
in the next patch with reasoning? It probably isn't safe if the lock is
taken in interrupt context anyway.

Yes, the drv->lock should have been irqsave/irqrestore, but it hasn't
been changed by this patch.


Signed-off-by: Raju P.L.S.S.S.N <rplsssn@xxxxxxxxxxxxxx>
[ilina: split patch into multiple files, update commit text]
Signed-off-by: Lina Iyer <ilina@xxxxxxxxxxxxxx>

diff --git a/drivers/soc/qcom/rpmh-internal.h b/drivers/soc/qcom/rpmh-internal.h
index a7bbbb67991c..969d5030860e 100644
--- a/drivers/soc/qcom/rpmh-internal.h
+++ b/drivers/soc/qcom/rpmh-internal.h
diff --git a/drivers/soc/qcom/rpmh-rsc.c b/drivers/soc/qcom/rpmh-rsc.c
index e278fc11fe5c..5ede8d6de3ad 100644
--- a/drivers/soc/qcom/rpmh-rsc.c
+++ b/drivers/soc/qcom/rpmh-rsc.c
@@ -106,26 +106,26 @@ static int tcs_invalidate(struct rsc_drv *drv, int type)
 {
        int m;
        struct tcs_group *tcs;
+       int ret = 0;

        tcs = get_tcs_of_type(drv, type);

-       spin_lock(&tcs->lock);
-       if (bitmap_empty(tcs->slots, MAX_TCS_SLOTS)) {
-               spin_unlock(&tcs->lock);
-               return 0;
-       }
+       spin_lock(&drv->lock);
+       if (bitmap_empty(tcs->slots, MAX_TCS_SLOTS))
+               goto done_invalidate;

        for (m = tcs->offset; m < tcs->offset + tcs->num_tcs; m++) {
                if (!tcs_is_free(drv, m)) {
-                       spin_unlock(&tcs->lock);
-                       return -EAGAIN;
+                       ret = -EAGAIN;
+                       goto done_invalidate;
                }
                write_tcs_reg_sync(drv, RSC_DRV_CMD_ENABLE, m, 0);
                write_tcs_reg_sync(drv, RSC_DRV_CMD_WAIT_FOR_CMPL, m, 0);
        }
        bitmap_zero(tcs->slots, MAX_TCS_SLOTS);
-       spin_unlock(&tcs->lock);

+done_invalidate:
+       spin_unlock(&drv->lock);
        return 0;

return ret now?

Yes, will do.
 }

@@ -349,41 +349,35 @@ static int tcs_write(struct rsc_drv *drv, const struct tcs_request *msg)
 {
        struct tcs_group *tcs;
        int tcs_id;
-       unsigned long flags;
        int ret;

        tcs = get_tcs_for_msg(drv, msg);
        if (IS_ERR(tcs))
                return PTR_ERR(tcs);

-       spin_lock_irqsave(&tcs->lock, flags);
        spin_lock(&drv->lock);
        /*
         * The h/w does not like if we send a request to the same address,
         * when one is already in-flight or being processed.
         */
        ret = check_for_req_inflight(drv, tcs, msg);
-       if (ret) {
-               spin_unlock(&drv->lock);
+       if (ret)
                goto done_write;
-       }

        tcs_id = find_free_tcs(tcs);
        if (tcs_id < 0) {
                ret = tcs_id;
-               spin_unlock(&drv->lock);
                goto done_write;
        }

        tcs->req[tcs_id - tcs->offset] = msg;
        set_bit(tcs_id, drv->tcs_in_use);
-       spin_unlock(&drv->lock);

        __tcs_buffer_write(drv, tcs_id, 0, msg);
        __tcs_trigger(drv, tcs_id);

 done_write:
-       spin_unlock_irqrestore(&tcs->lock, flags);
+       spin_unlock(&drv->lock);
        return ret;
 }

@@ -481,19 +475,18 @@ static int tcs_ctrl_write(struct rsc_drv *drv, const struct tcs_request *msg)
 {
        struct tcs_group *tcs;
        int tcs_id = 0, cmd_id = 0;
-       unsigned long flags;
        int ret;

        tcs = get_tcs_for_msg(drv, msg);
        if (IS_ERR(tcs))
                return PTR_ERR(tcs);

-       spin_lock_irqsave(&tcs->lock, flags);
+       spin_lock(&drv->lock);
        /* find the TCS id and the command in the TCS to write to */
        ret = find_slots(tcs, msg, &tcs_id, &cmd_id);
        if (!ret)
                __tcs_buffer_write(drv, tcs_id, cmd_id, msg);
-       spin_unlock_irqrestore(&tcs->lock, flags);
+       spin_unlock(&drv->lock);


These ones, just leave them doing the irq save restore for now?

drv->lock ??

--Lina




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux