| Message ID | 1502903951-5403-14-git-send-email-sricharan@codeaurora.org (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
On 8/16/2017 10:49 PM, Sricharan R wrote: > Send RX data receive ack to remote and also inform > that local intent buffer is used and freed. This > informs the remote to request for next set of > intent buffers before doing a send operation. > > Signed-off-by: Sricharan R <sricharan@codeaurora.org> > Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org> > --- > drivers/rpmsg/qcom_glink_native.c | 83 +++++++++++++++++++++++++++++++++++++++ > 1 file changed, 83 insertions(+) > > diff --git a/drivers/rpmsg/qcom_glink_native.c b/drivers/rpmsg/qcom_glink_native.c > index b2a583a..b8db74a 100644 > --- a/drivers/rpmsg/qcom_glink_native.c > +++ b/drivers/rpmsg/qcom_glink_native.c > @@ -76,6 +76,8 @@ struct glink_core_rx_intent { > bool reuse; > bool in_use; > u32 offset; > + > + struct list_head node; > }; > > /** > @@ -137,6 +139,8 @@ enum { > * @rcid: channel id, in remote space > * @intent_lock: lock for protection of @liids > * @liids: idr of all local intents > + * @intent_work: worker responsible for transmitting rx_done packets > + * @done_intents: list of intents that needs to be announced rx_done > * @buf: receive buffer, for gathering fragments > * @buf_offset: write offset in @buf > * @buf_size: size of current @buf > @@ -159,6 +163,8 @@ struct glink_channel { > > spinlock_t intent_lock; > struct idr liids; > + struct work_struct intent_work; > + struct list_head done_intents; > > struct glink_core_rx_intent *buf; > int buf_offset; > @@ -178,15 +184,19 @@ struct glink_channel { > #define RPM_CMD_CLOSE 3 > #define RPM_CMD_OPEN_ACK 4 > #define RPM_CMD_INTENT 5 > +#define RPM_CMD_RX_DONE 6 > #define RPM_CMD_RX_INTENT_REQ 7 > #define RPM_CMD_RX_INTENT_REQ_ACK 8 > #define RPM_CMD_TX_DATA 9 > #define RPM_CMD_CLOSE_ACK 11 > #define RPM_CMD_TX_DATA_CONT 12 > #define RPM_CMD_READ_NOTIF 13 > +#define RPM_CMD_RX_DONE_W_REUSE 14 > > #define GLINK_FEATURE_INTENTLESS BIT(1) > > +static void qcom_glink_rx_done_work(struct work_struct *work); > + > static struct glink_channel *qcom_glink_alloc_channel(struct qcom_glink *glink, > const char *name) > { > @@ -198,12 +208,16 @@ static struct glink_channel *qcom_glink_alloc_channel(struct qcom_glink *glink, > > /* Setup glink internal glink_channel data */ > spin_lock_init(&channel->recv_lock); > + > channel->glink = glink; > channel->name = kstrdup(name, GFP_KERNEL); > > init_completion(&channel->open_req); > init_completion(&channel->open_ack); > > + INIT_LIST_HEAD(&channel->done_intents); > + INIT_WORK(&channel->intent_work, qcom_glink_rx_done_work); > + > idr_init(&channel->liids); > kref_init(&channel->refcount); > > @@ -395,6 +409,70 @@ static void qcom_glink_send_close_ack(struct qcom_glink *glink, > qcom_glink_tx(glink, &req, sizeof(req), NULL, 0, true); > } > > +static void qcom_glink_rx_done_work(struct work_struct *work) > +{ > + struct glink_channel *channel = container_of(work, struct glink_channel, > + intent_work); > + struct qcom_glink *glink = channel->glink; > + struct glink_core_rx_intent *intent, *tmp; > + struct { > + u16 id; > + u16 lcid; > + u32 liid; > + } __packed cmd; > + > + unsigned int cid = channel->lcid; > + unsigned int iid; > + bool reuse; > + unsigned long flags; > + > + spin_lock_irqsave(&channel->intent_lock, flags); > + list_for_each_entry_safe(intent, tmp, &channel->done_intents, node) { > + list_del(&intent->node); > + spin_unlock_irqrestore(&channel->intent_lock, flags); > + iid = intent->id; > + reuse = intent->reuse; > + > + cmd.id = reuse ? RPM_CMD_RX_DONE_W_REUSE : RPM_CMD_RX_DONE; > + cmd.lcid = cid; > + cmd.liid = iid; > + > + qcom_glink_tx(glink, &cmd, sizeof(cmd), NULL, 0, true); > + if (!reuse) { > + kfree(intent->data); > + kfree(intent); > + } > + spin_lock_irqsave(&channel->intent_lock, flags); > + } > + spin_unlock_irqrestore(&channel->intent_lock, flags); > +} > + > +static void qcom_glink_rx_done(struct qcom_glink *glink, > + struct glink_channel *channel, > + struct glink_core_rx_intent *intent) > +{ > + /* We don't send RX_DONE to intentless systems */ > + if (glink->intentless) { > + kfree(intent->data); > + kfree(intent); > + return; > + } > + > + /* Take it off the tree of receive intents */ > + if (!intent->reuse) { > + spin_lock(&channel->intent_lock); > + idr_remove(&channel->liids, intent->id); > + spin_unlock(&channel->intent_lock); > + } > + > + /* Schedule the sending of a rx_done indication */ > + spin_lock(&channel->intent_lock); > + list_add_tail(&intent->node, &channel->done_intents); > + spin_unlock(&channel->intent_lock); > + > + schedule_work(&channel->intent_work); Adding one more parallel path will hit performance, if this worker could not get CPU cycles or blocked by other RT or HIGH_PRIO worker on global worker pool. > +} > + > /** > * qcom_glink_receive_version() - receive version/features from remote system > * > @@ -711,6 +789,8 @@ static int qcom_glink_rx_data(struct qcom_glink *glink, size_t avail) > > intent->offset = 0; > channel->buf = NULL; > + > + qcom_glink_rx_done(glink, channel, intent); > } > > advance_rx: > @@ -1100,6 +1180,9 @@ static void qcom_glink_rx_close(struct qcom_glink *glink, unsigned int rcid) > if (WARN(!channel, "close request on unknown channel\n")) > return; > > + /* cancel pending rx_done work */ > + cancel_work_sync(&channel->intent_work); > + > if (channel->rpdev) { > strncpy(chinfo.name, channel->name, sizeof(chinfo.name)); > chinfo.src = RPMSG_ADDR_ANY; -- To unsubscribe from this list: send the line "unsubscribe linux-remoteproc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, >> + /* Take it off the tree of receive intents */ >> + if (!intent->reuse) { >> + spin_lock(&channel->intent_lock); >> + idr_remove(&channel->liids, intent->id); >> + spin_unlock(&channel->intent_lock); >> + } >> + >> + /* Schedule the sending of a rx_done indication */ >> + spin_lock(&channel->intent_lock); >> + list_add_tail(&intent->node, &channel->done_intents); >> + spin_unlock(&channel->intent_lock); >> + >> + schedule_work(&channel->intent_work); > > Adding one more parallel path will hit performance, if this worker could not get CPU cycles > or blocked by other RT or HIGH_PRIO worker on global worker pool. The idea is, by design to have parallel non-blocking paths for rx and tx (that is done as a part of rx by sending the rx_done command), otherwise trying to send the rx_done command in the rx isr context is a problem since the tx can wait for the FIFO space and in worst case, can even lead to a potential deadlock if both the local and remote try the same. Having said that, instead of queuing this work in to the global queue, this can be put in to a local glink edge owned queue (or) a threaded isr ?, downstream does the rx_done in a client specific worker. Regards, Sricharan
On 8/22/2017 7:46 PM, Sricharan R wrote: > Hi, >>> + /* Take it off the tree of receive intents */ >>> + if (!intent->reuse) { >>> + spin_lock(&channel->intent_lock); >>> + idr_remove(&channel->liids, intent->id); >>> + spin_unlock(&channel->intent_lock); >>> + } >>> + >>> + /* Schedule the sending of a rx_done indication */ >>> + spin_lock(&channel->intent_lock); >>> + list_add_tail(&intent->node, &channel->done_intents); >>> + spin_unlock(&channel->intent_lock); >>> + >>> + schedule_work(&channel->intent_work); >> Adding one more parallel path will hit performance, if this worker could not get CPU cycles >> or blocked by other RT or HIGH_PRIO worker on global worker pool. > The idea is, by design to have parallel non-blocking paths for rx and tx (that is done as a > part of rx by sending the rx_done command), otherwise trying to send the rx_done > command in the rx isr context is a problem since the tx can wait for the FIFO space and > in worst case, can even lead to a potential deadlock if both the local and remote try > the same. Having said that, instead of queuing this work in to the global queue, this > can be put in to a local glink edge owned queue (or) a threaded isr ?, downstream does the > rx_done in a client specific worker. Yes, mixing RX and TX path will cause dead lock. I am okay to use specific queue with HIGH_PRIO or a threaded isr. down stream uses both client specific worker and client RX cb [this mix the TX and RX path] which want to avoid. > > Regards, > Sricharan > -- To unsubscribe from this list: send the line "unsubscribe linux-remoteproc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/rpmsg/qcom_glink_native.c b/drivers/rpmsg/qcom_glink_native.c index b2a583a..b8db74a 100644 --- a/drivers/rpmsg/qcom_glink_native.c +++ b/drivers/rpmsg/qcom_glink_native.c @@ -76,6 +76,8 @@ struct glink_core_rx_intent { bool reuse; bool in_use; u32 offset; + + struct list_head node; }; /** @@ -137,6 +139,8 @@ enum { * @rcid: channel id, in remote space * @intent_lock: lock for protection of @liids * @liids: idr of all local intents + * @intent_work: worker responsible for transmitting rx_done packets + * @done_intents: list of intents that needs to be announced rx_done * @buf: receive buffer, for gathering fragments * @buf_offset: write offset in @buf * @buf_size: size of current @buf @@ -159,6 +163,8 @@ struct glink_channel { spinlock_t intent_lock; struct idr liids; + struct work_struct intent_work; + struct list_head done_intents; struct glink_core_rx_intent *buf; int buf_offset; @@ -178,15 +184,19 @@ struct glink_channel { #define RPM_CMD_CLOSE 3 #define RPM_CMD_OPEN_ACK 4 #define RPM_CMD_INTENT 5 +#define RPM_CMD_RX_DONE 6 #define RPM_CMD_RX_INTENT_REQ 7 #define RPM_CMD_RX_INTENT_REQ_ACK 8 #define RPM_CMD_TX_DATA 9 #define RPM_CMD_CLOSE_ACK 11 #define RPM_CMD_TX_DATA_CONT 12 #define RPM_CMD_READ_NOTIF 13 +#define RPM_CMD_RX_DONE_W_REUSE 14 #define GLINK_FEATURE_INTENTLESS BIT(1) +static void qcom_glink_rx_done_work(struct work_struct *work); + static struct glink_channel *qcom_glink_alloc_channel(struct qcom_glink *glink, const char *name) { @@ -198,12 +208,16 @@ static struct glink_channel *qcom_glink_alloc_channel(struct qcom_glink *glink, /* Setup glink internal glink_channel data */ spin_lock_init(&channel->recv_lock); + channel->glink = glink; channel->name = kstrdup(name, GFP_KERNEL); init_completion(&channel->open_req); init_completion(&channel->open_ack); + INIT_LIST_HEAD(&channel->done_intents); + INIT_WORK(&channel->intent_work, qcom_glink_rx_done_work); + idr_init(&channel->liids); kref_init(&channel->refcount); @@ -395,6 +409,70 @@ static void qcom_glink_send_close_ack(struct qcom_glink *glink, qcom_glink_tx(glink, &req, sizeof(req), NULL, 0, true); } +static void qcom_glink_rx_done_work(struct work_struct *work) +{ + struct glink_channel *channel = container_of(work, struct glink_channel, + intent_work); + struct qcom_glink *glink = channel->glink; + struct glink_core_rx_intent *intent, *tmp; + struct { + u16 id; + u16 lcid; + u32 liid; + } __packed cmd; + + unsigned int cid = channel->lcid; + unsigned int iid; + bool reuse; + unsigned long flags; + + spin_lock_irqsave(&channel->intent_lock, flags); + list_for_each_entry_safe(intent, tmp, &channel->done_intents, node) { + list_del(&intent->node); + spin_unlock_irqrestore(&channel->intent_lock, flags); + iid = intent->id; + reuse = intent->reuse; + + cmd.id = reuse ? RPM_CMD_RX_DONE_W_REUSE : RPM_CMD_RX_DONE; + cmd.lcid = cid; + cmd.liid = iid; + + qcom_glink_tx(glink, &cmd, sizeof(cmd), NULL, 0, true); + if (!reuse) { + kfree(intent->data); + kfree(intent); + } + spin_lock_irqsave(&channel->intent_lock, flags); + } + spin_unlock_irqrestore(&channel->intent_lock, flags); +} + +static void qcom_glink_rx_done(struct qcom_glink *glink, + struct glink_channel *channel, + struct glink_core_rx_intent *intent) +{ + /* We don't send RX_DONE to intentless systems */ + if (glink->intentless) { + kfree(intent->data); + kfree(intent); + return; + } + + /* Take it off the tree of receive intents */ + if (!intent->reuse) { + spin_lock(&channel->intent_lock); + idr_remove(&channel->liids, intent->id); + spin_unlock(&channel->intent_lock); + } + + /* Schedule the sending of a rx_done indication */ + spin_lock(&channel->intent_lock); + list_add_tail(&intent->node, &channel->done_intents); + spin_unlock(&channel->intent_lock); + + schedule_work(&channel->intent_work); +} + /** * qcom_glink_receive_version() - receive version/features from remote system * @@ -711,6 +789,8 @@ static int qcom_glink_rx_data(struct qcom_glink *glink, size_t avail) intent->offset = 0; channel->buf = NULL; + + qcom_glink_rx_done(glink, channel, intent); } advance_rx: @@ -1100,6 +1180,9 @@ static void qcom_glink_rx_close(struct qcom_glink *glink, unsigned int rcid) if (WARN(!channel, "close request on unknown channel\n")) return; + /* cancel pending rx_done work */ + cancel_work_sync(&channel->intent_work); + if (channel->rpdev) { strncpy(chinfo.name, channel->name, sizeof(chinfo.name)); chinfo.src = RPMSG_ADDR_ANY;