<p>laforge <strong>submitted</strong> this change.</p><p><a href="https://gerrit.osmocom.org/c/osmo-e1d/+/19345">View Change</a></p><div style="white-space:pre-wrap">Approvals:
Jenkins Builder: Verified
laforge: Looks good to me, approved
</div><pre style="font-family: monospace,monospace; white-space: pre-wrap;">implement buffering to chunks in E1->application direction<br><br>Existing applications (such as those written for DAHDI) expect to<br>be reading data in buffer/chunk sizes. For example OsmoNITB: It<br>doesn't want to execute an expensive read/recv syscall to receive 11<br>bytes, if it needs at least 160 bytes.<br><br>Change-Id: I807671bc6f2acaef740ce215b8d8abcb5dce2640<br>---<br>M src/ctl.c<br>M src/e1d.h<br>M src/intf_line.c<br>3 files changed, 66 insertions(+), 2 deletions(-)<br><br></pre><pre style="font-family: monospace,monospace; white-space: pre-wrap;"><span>diff --git a/src/ctl.c b/src/ctl.c</span><br><span>index 697af01..d62195a 100644</span><br><span>--- a/src/ctl.c</span><br><span>+++ b/src/ctl.c</span><br><span>@@ -121,6 +121,20 @@</span><br><span> close(ts->fd);</span><br><span> ts->fd = -1;</span><br><span> }</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ talloc_free(ts->raw.rx_buf);</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf = NULL;</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf_size = 0;</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf_used = 0;</span><br><span style="color: hsl(120, 100%, 40%);">+}</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+static void</span><br><span style="color: hsl(120, 100%, 40%);">+_e1d_ts_raw_buf_realloc(struct e1_ts *ts, unsigned int size)</span><br><span style="color: hsl(120, 100%, 40%);">+{</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf = talloc_realloc_size(ts->line, ts->raw.rx_buf, size);</span><br><span style="color: hsl(120, 100%, 40%);">+ OSMO_ASSERT(ts->raw.rx_buf);</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf_size = size;</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf_used = 0;</span><br><span> }</span><br><span> </span><br><span> static int</span><br><span>@@ -137,6 +151,7 @@</span><br><span> break;</span><br><span> case E1_TS_MODE_RAW:</span><br><span> sock_type = SOCK_STREAM;</span><br><span style="color: hsl(120, 100%, 40%);">+ _e1d_ts_raw_buf_realloc(ts, 160); /* TODO: make configurable */</span><br><span> break;</span><br><span> default:</span><br><span> return -EINVAL;</span><br><span>diff --git a/src/e1d.h b/src/e1d.h</span><br><span>index 98c3ddc..c35f4a8 100644</span><br><span>--- a/src/e1d.h</span><br><span>+++ b/src/e1d.h</span><br><span>@@ -53,6 +53,13 @@</span><br><span> int tx_len;</span><br><span> } hdlc;</span><br><span> </span><br><span style="color: hsl(120, 100%, 40%);">+ /* RAW handling */</span><br><span style="color: hsl(120, 100%, 40%);">+ struct {</span><br><span style="color: hsl(120, 100%, 40%);">+ uint8_t *rx_buf; /* actual buffer storage */</span><br><span style="color: hsl(120, 100%, 40%);">+ unsigned int rx_buf_size; /* size of 'buf' in bytes */</span><br><span style="color: hsl(120, 100%, 40%);">+ unsigned int rx_buf_used; /* number of bytes used so far */</span><br><span style="color: hsl(120, 100%, 40%);">+ } raw;</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span> /* Remote end */</span><br><span> int fd;</span><br><span> };</span><br><span>diff --git a/src/intf_line.c b/src/intf_line.c</span><br><span>index 6213151..5fd82b7 100644</span><br><span>--- a/src/intf_line.c</span><br><span>+++ b/src/intf_line.c</span><br><span>@@ -200,7 +200,7 @@</span><br><span> }</span><br><span> LOGPTS(ts, DXFR, LOGL_DEBUG, "TX Message: %d [ %s]\n",</span><br><span> rv, osmo_hexdump(ts->hdlc.tx_buf, rv));</span><br><span style="color: hsl(0, 100%, 40%);">- ts->hdlc.tx_len = rv; </span><br><span style="color: hsl(120, 100%, 40%);">+ ts->hdlc.tx_len = rv;</span><br><span> ts->hdlc.tx_ofs = 0;</span><br><span> } else if (rv < 0 && errno != EAGAIN)</span><br><span> return rv;</span><br><span>@@ -338,6 +338,48 @@</span><br><span> return tsz;</span><br><span> }</span><br><span> </span><br><span style="color: hsl(120, 100%, 40%);">+/* append data to the per-timeslot buffer; flush to socket every time buffer is full */</span><br><span style="color: hsl(120, 100%, 40%);">+static int</span><br><span style="color: hsl(120, 100%, 40%);">+_e1_rx_raw(struct e1_ts *ts, const uint8_t *buf, unsigned int len)</span><br><span style="color: hsl(120, 100%, 40%);">+{</span><br><span style="color: hsl(120, 100%, 40%);">+ unsigned int appended = 0;</span><br><span style="color: hsl(120, 100%, 40%);">+ int rv;</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ OSMO_ASSERT(ts->mode == E1_TS_MODE_RAW);</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ /* we don't keep a larger set of buffers but simply assume that whenever</span><br><span style="color: hsl(120, 100%, 40%);">+ * we received one full chunk/buffer size, we are able to push the data</span><br><span style="color: hsl(120, 100%, 40%);">+ * into the underlying unix domain socket. Kernel socket buffering should</span><br><span style="color: hsl(120, 100%, 40%);">+ * be far sufficient in terms of buffering capacity of voice data (which</span><br><span style="color: hsl(120, 100%, 40%);">+ * is typically consumed reasonably low latency and hence buffer size) */</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ while (appended < len) {</span><br><span style="color: hsl(120, 100%, 40%);">+ unsigned int ts_buf_tailroom = ts->raw.rx_buf_size - ts->raw.rx_buf_used;</span><br><span style="color: hsl(120, 100%, 40%);">+ unsigned int chunk_len;</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ /* determine size of chunk we can write at this point */</span><br><span style="color: hsl(120, 100%, 40%);">+ chunk_len = len - appended;</span><br><span style="color: hsl(120, 100%, 40%);">+ if (chunk_len > ts_buf_tailroom)</span><br><span style="color: hsl(120, 100%, 40%);">+ chunk_len = ts_buf_tailroom;</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ /* actually copy the chunk */</span><br><span style="color: hsl(120, 100%, 40%);">+ memcpy(ts->raw.rx_buf + ts->raw.rx_buf_used, buf + appended, chunk_len);</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf_used += chunk_len;</span><br><span style="color: hsl(120, 100%, 40%);">+ appended += chunk_len;</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ /* if ts_buf is full: flush + rewind */</span><br><span style="color: hsl(120, 100%, 40%);">+ if (ts->raw.rx_buf_used >= ts->raw.rx_buf_size) {</span><br><span style="color: hsl(120, 100%, 40%);">+ rv = write(ts->fd, ts->raw.rx_buf, ts->raw.rx_buf_size);</span><br><span style="color: hsl(120, 100%, 40%);">+ if (rv < 0)</span><br><span style="color: hsl(120, 100%, 40%);">+ return rv;</span><br><span style="color: hsl(120, 100%, 40%);">+ /* FIXME: count overflows */</span><br><span style="color: hsl(120, 100%, 40%);">+ ts->raw.rx_buf_used = 0;</span><br><span style="color: hsl(120, 100%, 40%);">+ }</span><br><span style="color: hsl(120, 100%, 40%);">+ }</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span style="color: hsl(120, 100%, 40%);">+ return appended;</span><br><span style="color: hsl(120, 100%, 40%);">+}</span><br><span style="color: hsl(120, 100%, 40%);">+</span><br><span> /* write data to a timeslot (hardware -> application direction) */</span><br><span> static int</span><br><span> _e1_ts_write(struct e1_ts *ts, const uint8_t *buf, size_t len)</span><br><span>@@ -346,7 +388,7 @@</span><br><span> </span><br><span> switch (ts->mode) {</span><br><span> case E1_TS_MODE_RAW:</span><br><span style="color: hsl(0, 100%, 40%);">- rv = write(ts->fd, buf, len);</span><br><span style="color: hsl(120, 100%, 40%);">+ rv = _e1_rx_raw(ts, buf, len);</span><br><span> break;</span><br><span> case E1_TS_MODE_HDLCFCS:</span><br><span> rv = _e1_rx_hdlcfs(ts, buf, len);</span><br><span></span><br></pre><p>To view, visit <a href="https://gerrit.osmocom.org/c/osmo-e1d/+/19345">change 19345</a>. To unsubscribe, or for help writing mail filters, visit <a href="https://gerrit.osmocom.org/settings">settings</a>.</p><div itemscope itemtype="http://schema.org/EmailMessage"><div itemscope itemprop="action" itemtype="http://schema.org/ViewAction"><link itemprop="url" href="https://gerrit.osmocom.org/c/osmo-e1d/+/19345"/><meta itemprop="name" content="View Change"/></div></div>
<div style="display:none"> Gerrit-Project: osmo-e1d </div>
<div style="display:none"> Gerrit-Branch: master </div>
<div style="display:none"> Gerrit-Change-Id: I807671bc6f2acaef740ce215b8d8abcb5dce2640 </div>
<div style="display:none"> Gerrit-Change-Number: 19345 </div>
<div style="display:none"> Gerrit-PatchSet: 3 </div>
<div style="display:none"> Gerrit-Owner: laforge <laforge@osmocom.org> </div>
<div style="display:none"> Gerrit-Assignee: tnt <tnt@246tNt.com> </div>
<div style="display:none"> Gerrit-Reviewer: Jenkins Builder </div>
<div style="display:none"> Gerrit-Reviewer: dexter <pmaier@sysmocom.de> </div>
<div style="display:none"> Gerrit-Reviewer: laforge <laforge@osmocom.org> </div>
<div style="display:none"> Gerrit-Reviewer: tnt <tnt@246tNt.com> </div>
<div style="display:none"> Gerrit-CC: pespin <pespin@sysmocom.de> </div>
<div style="display:none"> Gerrit-MessageType: merged </div>