| Message ID | 20220628024913.1755292-6-tzungbi@kernel.org (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | platform/chrome: Kunit tests and refactor for cros_ec_cmd_xfer() | expand |
On Tue, Jun 28, 2022 at 02:49:07AM +0000, Tzung-Bi Shih wrote: > EC returns EC_RES_IN_PROGRESS if the host command needs more time to > complete. Whenever receives the return code, cros_ec_send_command() > sends EC_CMD_GET_COMMS_STATUS to query the command status. > > Separate cros_ec_wait_until_complete() from cros_ec_send_command(). > It sends EC_CMD_GET_COMMS_STATUS and waits until the previous command > was completed, or encountered error, or timed out. > > Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org> Could someone on the list help to review the remaining patches in the series?
On Mon, Jun 27, 2022 at 7:49 PM Tzung-Bi Shih <tzungbi@kernel.org> wrote: > > EC returns EC_RES_IN_PROGRESS if the host command needs more time to > complete. Whenever receives the return code, cros_ec_send_command() > sends EC_CMD_GET_COMMS_STATUS to query the command status. > > Separate cros_ec_wait_until_complete() from cros_ec_send_command(). > It sends EC_CMD_GET_COMMS_STATUS and waits until the previous command > was completed, or encountered error, or timed out. > > Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org> > --- > drivers/platform/chrome/cros_ec_proto.c | 75 ++++++++++++------------- > 1 file changed, 36 insertions(+), 39 deletions(-) > > diff --git a/drivers/platform/chrome/cros_ec_proto.c b/drivers/platform/chrome/cros_ec_proto.c > index 0cec013be3d3..466ecb063bd6 100644 > --- a/drivers/platform/chrome/cros_ec_proto.c > +++ b/drivers/platform/chrome/cros_ec_proto.c > @@ -134,53 +134,50 @@ static int cros_ec_xfer_command(struct cros_ec_device *ec_dev, struct cros_ec_co > return ret; > } > > -static int cros_ec_send_command(struct cros_ec_device *ec_dev, struct cros_ec_command *msg) > +static int cros_ec_wait_until_complete(struct cros_ec_device *ec_dev, uint32_t *result) > { > - int ret = cros_ec_xfer_command(ec_dev, msg); > + struct cros_ec_command *msg; > + struct ec_response_get_comms_status *status; > + int ret = 0, i; > + > + msg = kzalloc(sizeof(*msg) + sizeof(*status), GFP_KERNEL); > + if (!msg) > + return -ENOMEM; AFAICS this is always 24 bytes. I would suggest to allocate it on the stack to reduce overhead. > > - if (msg->result == EC_RES_IN_PROGRESS) { > - int i; > - struct cros_ec_command *status_msg; > - struct ec_response_get_comms_status *status; > + msg->command = EC_CMD_GET_COMMS_STATUS; > + msg->insize = sizeof(*status); > > - status_msg = kmalloc(sizeof(*status_msg) + sizeof(*status), > - GFP_KERNEL); > - if (!status_msg) > - return -ENOMEM; > + status = (struct ec_response_get_comms_status *)msg->data; > > - status_msg->version = 0; > - status_msg->command = EC_CMD_GET_COMMS_STATUS; > - status_msg->insize = sizeof(*status); > - status_msg->outsize = 0; > + /* Query the EC's status until it's no longer busy or we encounter an error. */ > + for (i = 0; i < EC_COMMAND_RETRIES; ++i) { > + usleep_range(10000, 11000); > > - /* > - * Query the EC's status until it's no longer busy or > - * we encounter an error. > - */ > - for (i = 0; i < EC_COMMAND_RETRIES; i++) { > - usleep_range(10000, 11000); > - > - trace_cros_ec_request_start(status_msg); > - ret = (*xfer_fxn)(ec_dev, status_msg); > - trace_cros_ec_request_done(status_msg, ret); > - if (ret == -EAGAIN) > - continue; > - if (ret < 0) > - break; > - > - msg->result = status_msg->result; > - if (status_msg->result != EC_RES_SUCCESS) > - break; > - > - status = (struct ec_response_get_comms_status *) > - status_msg->data; > - if (!(status->flags & EC_COMMS_STATUS_PROCESSING)) > - break; > - } > + ret = cros_ec_xfer_command(ec_dev, msg); > + if (ret == -EAGAIN) > + continue; > + if (ret < 0) > + break; With the command allocated on the stack, this can return immediately. > + > + *result = msg->result; > + if (msg->result != EC_RES_SUCCESS) > + break; Again, this can return immediately if the command buffer is on the stack. > > - kfree(status_msg); > + if (!(status->flags & EC_COMMS_STATUS_PROCESSING)) > + break; Can return immediately. > } > > + kfree(msg); > + return ret; What should this return on timeout ? > +} > + > +static int cros_ec_send_command(struct cros_ec_device *ec_dev, struct cros_ec_command *msg) > +{ > + int ret = cros_ec_xfer_command(ec_dev, msg); > + > + if (msg->result == EC_RES_IN_PROGRESS) > + ret = cros_ec_wait_until_complete(ec_dev, &msg->result); > + > return ret; > } > > -- > 2.37.0.rc0.161.g10f37bed90-goog >
On Wed, Jul 13, 2022 at 11:15:47AM -0700, Guenter Roeck wrote: > On Mon, Jun 27, 2022 at 7:49 PM Tzung-Bi Shih <tzungbi@kernel.org> wrote: > > -static int cros_ec_send_command(struct cros_ec_device *ec_dev, struct cros_ec_command *msg) > > +static int cros_ec_wait_until_complete(struct cros_ec_device *ec_dev, uint32_t *result) > > { > > - int ret = cros_ec_xfer_command(ec_dev, msg); > > + struct cros_ec_command *msg; > > + struct ec_response_get_comms_status *status; > > + int ret = 0, i; > > + > > + msg = kzalloc(sizeof(*msg) + sizeof(*status), GFP_KERNEL); > > + if (!msg) > > + return -ENOMEM; > > AFAICS this is always 24 bytes. I would suggest to allocate it on the > stack to reduce overhead. Ack. > > + ret = cros_ec_xfer_command(ec_dev, msg); > > + if (ret == -EAGAIN) > > + continue; > > + if (ret < 0) > > + break; > > With the command allocated on the stack, this can return immediately. Nack, the function has no goto labels. `return ret` follows the loop immediately. The `break` here doesn't make it to become too complicated. I would prefer to keep it. > > + > > + *result = msg->result; > > + if (msg->result != EC_RES_SUCCESS) > > + break; > > Again, this can return immediately if the command buffer is on the stack. Nack. See above. > > - kfree(status_msg); > > + if (!(status->flags & EC_COMMS_STATUS_PROCESSING)) > > + break; > > Can return immediately. Nack. See above. > > + kfree(msg); > > + return ret; > > What should this return on timeout ? It returns either: * -EAGAIN, if cros_ec_xfer_command() returned -EAGAIN * 0, if EC_COMMS_STATUS_PROCESSING flag was on for EC_COMMAND_RETRIES times so far. This is a "move" refactor. I would prefer to keep it as is and change the behavior in later patch.
On Wed, Jul 13, 2022 at 8:33 PM Tzung-Bi Shih <tzungbi@kernel.org> wrote: > > On Wed, Jul 13, 2022 at 11:15:47AM -0700, Guenter Roeck wrote: > > On Mon, Jun 27, 2022 at 7:49 PM Tzung-Bi Shih <tzungbi@kernel.org> wrote: > > > -static int cros_ec_send_command(struct cros_ec_device *ec_dev, struct cros_ec_command *msg) > > > +static int cros_ec_wait_until_complete(struct cros_ec_device *ec_dev, uint32_t *result) > > > { > > > - int ret = cros_ec_xfer_command(ec_dev, msg); > > > + struct cros_ec_command *msg; > > > + struct ec_response_get_comms_status *status; > > > + int ret = 0, i; > > > + > > > + msg = kzalloc(sizeof(*msg) + sizeof(*status), GFP_KERNEL); > > > + if (!msg) > > > + return -ENOMEM; > > > > AFAICS this is always 24 bytes. I would suggest to allocate it on the > > stack to reduce overhead. > > Ack. > > > > + ret = cros_ec_xfer_command(ec_dev, msg); > > > + if (ret == -EAGAIN) > > > + continue; > > > + if (ret < 0) > > > + break; > > > > With the command allocated on the stack, this can return immediately. > > Nack, the function has no goto labels. `return ret` follows the loop > immediately. The `break` here doesn't make it to become too complicated. > I would prefer to keep it. Sorry, you lost me here. The code after the loop does kfree(msg); return ret; If kfree() is no longer necessary, only the return statement is left. So break; is identical to return ret;. Am I missing something ? > > > > + > > > + *result = msg->result; > > > + if (msg->result != EC_RES_SUCCESS) > > > + break; > > > > Again, this can return immediately if the command buffer is on the stack. > > Nack. See above. > > > > - kfree(status_msg); > > > + if (!(status->flags & EC_COMMS_STATUS_PROCESSING)) > > > + break; > > > > Can return immediately. > > Nack. See above. > Really, for those I think that return 0; would be better and more explicit. > > > + kfree(msg); > > > + return ret; > > > > What should this return on timeout ? > > It returns either: > * -EAGAIN, if cros_ec_xfer_command() returned -EAGAIN > * 0, if EC_COMMS_STATUS_PROCESSING flag was on > for EC_COMMAND_RETRIES times so far. > > This is a "move" refactor. I would prefer to keep it as is and change the > behavior in later patch. Ok. Thanks, Guenter
On Thu, Jul 14, 2022 at 07:29:45AM -0700, Guenter Roeck wrote: > On Wed, Jul 13, 2022 at 8:33 PM Tzung-Bi Shih <tzungbi@kernel.org> wrote: > > On Wed, Jul 13, 2022 at 11:15:47AM -0700, Guenter Roeck wrote: > > > On Mon, Jun 27, 2022 at 7:49 PM Tzung-Bi Shih <tzungbi@kernel.org> wrote: > > > > + ret = cros_ec_xfer_command(ec_dev, msg); > > > > + if (ret == -EAGAIN) > > > > + continue; > > > > + if (ret < 0) > > > > + break; > > > > > > With the command allocated on the stack, this can return immediately. > > > > Nack, the function has no goto labels. `return ret` follows the loop > > immediately. The `break` here doesn't make it to become too complicated. > > I would prefer to keep it. > > Sorry, you lost me here. The code after the loop does > > kfree(msg); > return ret; > > If kfree() is no longer necessary, only the return statement is left. So break; > is identical to return ret;. Am I missing something ? You are correct. I meant personally I would prefer to use `break`: * The loop is short so that it won't become too complicated. * Keep the function has a single exit point. But, anyway, let's use `return ret` to make it explicit.
diff --git a/drivers/platform/chrome/cros_ec_proto.c b/drivers/platform/chrome/cros_ec_proto.c index 0cec013be3d3..466ecb063bd6 100644 --- a/drivers/platform/chrome/cros_ec_proto.c +++ b/drivers/platform/chrome/cros_ec_proto.c @@ -134,53 +134,50 @@ static int cros_ec_xfer_command(struct cros_ec_device *ec_dev, struct cros_ec_co return ret; } -static int cros_ec_send_command(struct cros_ec_device *ec_dev, struct cros_ec_command *msg) +static int cros_ec_wait_until_complete(struct cros_ec_device *ec_dev, uint32_t *result) { - int ret = cros_ec_xfer_command(ec_dev, msg); + struct cros_ec_command *msg; + struct ec_response_get_comms_status *status; + int ret = 0, i; + + msg = kzalloc(sizeof(*msg) + sizeof(*status), GFP_KERNEL); + if (!msg) + return -ENOMEM; - if (msg->result == EC_RES_IN_PROGRESS) { - int i; - struct cros_ec_command *status_msg; - struct ec_response_get_comms_status *status; + msg->command = EC_CMD_GET_COMMS_STATUS; + msg->insize = sizeof(*status); - status_msg = kmalloc(sizeof(*status_msg) + sizeof(*status), - GFP_KERNEL); - if (!status_msg) - return -ENOMEM; + status = (struct ec_response_get_comms_status *)msg->data; - status_msg->version = 0; - status_msg->command = EC_CMD_GET_COMMS_STATUS; - status_msg->insize = sizeof(*status); - status_msg->outsize = 0; + /* Query the EC's status until it's no longer busy or we encounter an error. */ + for (i = 0; i < EC_COMMAND_RETRIES; ++i) { + usleep_range(10000, 11000); - /* - * Query the EC's status until it's no longer busy or - * we encounter an error. - */ - for (i = 0; i < EC_COMMAND_RETRIES; i++) { - usleep_range(10000, 11000); - - trace_cros_ec_request_start(status_msg); - ret = (*xfer_fxn)(ec_dev, status_msg); - trace_cros_ec_request_done(status_msg, ret); - if (ret == -EAGAIN) - continue; - if (ret < 0) - break; - - msg->result = status_msg->result; - if (status_msg->result != EC_RES_SUCCESS) - break; - - status = (struct ec_response_get_comms_status *) - status_msg->data; - if (!(status->flags & EC_COMMS_STATUS_PROCESSING)) - break; - } + ret = cros_ec_xfer_command(ec_dev, msg); + if (ret == -EAGAIN) + continue; + if (ret < 0) + break; + + *result = msg->result; + if (msg->result != EC_RES_SUCCESS) + break; - kfree(status_msg); + if (!(status->flags & EC_COMMS_STATUS_PROCESSING)) + break; } + kfree(msg); + return ret; +} + +static int cros_ec_send_command(struct cros_ec_device *ec_dev, struct cros_ec_command *msg) +{ + int ret = cros_ec_xfer_command(ec_dev, msg); + + if (msg->result == EC_RES_IN_PROGRESS) + ret = cros_ec_wait_until_complete(ec_dev, &msg->result); + return ret; }
EC returns EC_RES_IN_PROGRESS if the host command needs more time to complete. Whenever receives the return code, cros_ec_send_command() sends EC_CMD_GET_COMMS_STATUS to query the command status. Separate cros_ec_wait_until_complete() from cros_ec_send_command(). It sends EC_CMD_GET_COMMS_STATUS and waits until the previous command was completed, or encountered error, or timed out. Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org> --- drivers/platform/chrome/cros_ec_proto.c | 75 ++++++++++++------------- 1 file changed, 36 insertions(+), 39 deletions(-)