From: Mathias Nyman <mathias.nyman@linux.intel.com>
To: <gregkh@linuxfoundation.org>
Cc: linux-usb@vger.kernel.org,
"Mathias Nyman" <mathias.nyman@linux.intel.com>,
"Łukasz Bartosik" <ukaszb@chromium.org>
Subject: [PATCH 2/5] xhci: dbgtty: Improve performance by handling received data immediately.
Date: Fri, 27 Dec 2024 14:01:39 +0200 [thread overview]
Message-ID: <20241227120142.1035206-3-mathias.nyman@linux.intel.com> (raw)
In-Reply-To: <20241227120142.1035206-1-mathias.nyman@linux.intel.com>
Improve dbc transfer rate performance by copying the received data to
the tty buffer directly in the request complete callback function if
possible.
Only defer it in case there is already pending deferred work, tty is
throttled, or we fail copy the data to the tty buffer
The request complete callback is already called by a workqueue.
This is part 3/3 of a dbc performance improvement series that roughly
triples dbc performace when using adb push and pull over dbc.
Max/min push rate after patches is 210/118 MB/s, pull rate 171/133 MB/s,
tested with large files (300MB-9GB) by Łukasz Bartosik
Cc: Łukasz Bartosik <ukaszb@chromium.org>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
---
drivers/usb/host/xhci-dbgtty.c | 98 ++++++++++++++++++++++------------
1 file changed, 65 insertions(+), 33 deletions(-)
diff --git a/drivers/usb/host/xhci-dbgtty.c b/drivers/usb/host/xhci-dbgtty.c
index d719c16ea30b..60ed753c85bb 100644
--- a/drivers/usb/host/xhci-dbgtty.c
+++ b/drivers/usb/host/xhci-dbgtty.c
@@ -110,15 +110,74 @@ static void dbc_start_rx(struct dbc_port *port)
}
}
+/*
+ * Queue received data to tty buffer and push it.
+ *
+ * Returns nr of remaining bytes that didn't fit tty buffer, i.e. 0 if all
+ * bytes sucessfullt moved. In case of error returns negative errno.
+ * Call with lock held
+ */
+static int dbc_rx_push_buffer(struct dbc_port *port, struct dbc_request *req)
+{
+ char *packet = req->buf;
+ unsigned int n, size = req->actual;
+ int count;
+
+ if (!req->actual)
+ return 0;
+
+ /* if n_read is set then request was partially moved to tty buffer */
+ n = port->n_read;
+ if (n) {
+ packet += n;
+ size -= n;
+ }
+
+ count = tty_insert_flip_string(&port->port, packet, size);
+ if (count)
+ tty_flip_buffer_push(&port->port);
+ if (count != size) {
+ port->n_read += count;
+ return size - count;
+ }
+
+ port->n_read = 0;
+ return 0;
+}
+
static void
dbc_read_complete(struct xhci_dbc *dbc, struct dbc_request *req)
{
unsigned long flags;
struct dbc_port *port = dbc_to_port(dbc);
+ struct tty_struct *tty;
+ int untransferred;
+
+ tty = port->port.tty;
spin_lock_irqsave(&port->port_lock, flags);
+
+ /*
+ * Only defer copyig data to tty buffer in case:
+ * - !list_empty(&port->read_queue), there are older pending data
+ * - tty is throttled
+ * - failed to copy all data to buffer, defer remaining part
+ */
+
+ if (list_empty(&port->read_queue) && tty && !tty_throttled(tty)) {
+ untransferred = dbc_rx_push_buffer(port, req);
+ if (untransferred == 0) {
+ list_add_tail(&req->list_pool, &port->read_pool);
+ if (req->status != -ESHUTDOWN)
+ dbc_start_rx(port);
+ goto out;
+ }
+ }
+
+ /* defer moving data from req to tty buffer to a tasklet */
list_add_tail(&req->list_pool, &port->read_queue);
tasklet_schedule(&port->push);
+out:
spin_unlock_irqrestore(&port->port_lock, flags);
}
@@ -331,10 +390,10 @@ static void dbc_rx_push(struct tasklet_struct *t)
struct dbc_request *req;
struct tty_struct *tty;
unsigned long flags;
- bool do_push = false;
bool disconnect = false;
struct dbc_port *port = from_tasklet(port, t, push);
struct list_head *queue = &port->read_queue;
+ int untransferred;
spin_lock_irqsave(&port->port_lock, flags);
tty = port->port.tty;
@@ -356,42 +415,15 @@ static void dbc_rx_push(struct tasklet_struct *t)
break;
}
- if (req->actual) {
- char *packet = req->buf;
- unsigned int n, size = req->actual;
- int count;
-
- n = port->n_read;
- if (n) {
- packet += n;
- size -= n;
- }
-
- count = tty_insert_flip_string(&port->port, packet,
- size);
- if (count)
- do_push = true;
- if (count != size) {
- port->n_read += count;
- break;
- }
- port->n_read = 0;
- }
+ untransferred = dbc_rx_push_buffer(port, req);
+ if (untransferred > 0)
+ break;
list_move_tail(&req->list_pool, &port->read_pool);
}
- if (do_push)
- tty_flip_buffer_push(&port->port);
-
- if (!list_empty(queue) && tty) {
- if (!tty_throttled(tty)) {
- if (do_push)
- tasklet_schedule(&port->push);
- else
- pr_warn("ttyDBC0: RX not scheduled?\n");
- }
- }
+ if (!list_empty(queue))
+ tasklet_schedule(&port->push);
if (!disconnect)
dbc_start_rx(port);
--
2.25.1
next prev parent reply other threads:[~2024-12-27 12:00 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-12-27 12:01 [PATCH 0/5] xhci features for usb-next Mathias Nyman
2024-12-27 12:01 ` [PATCH 1/5] xhci: dbc: Improve performance by removing delay in transfer event polling Mathias Nyman
2024-12-27 12:01 ` Mathias Nyman [this message]
2024-12-27 12:01 ` [PATCH 3/5] usb: xhci: Fix NULL pointer dereference on certain command aborts Mathias Nyman
2024-12-27 12:01 ` [PATCH 4/5] xhci: Add command completion parameter support Mathias Nyman
2024-12-27 12:01 ` [PATCH 5/5] xhci: Add missing capability definition bits Mathias Nyman
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20241227120142.1035206-3-mathias.nyman@linux.intel.com \
--to=mathias.nyman@linux.intel.com \
--cc=gregkh@linuxfoundation.org \
--cc=linux-usb@vger.kernel.org \
--cc=ukaszb@chromium.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.