History log of /trafficserver/iocore/eventsystem/ (Results 101 - 125 of 447)
Revision (<<< Hide revision tags) (Show revision tags >>>)Date Author Comments
(<<< Hide modified files)
(Show modified files >>>)
20f1fda210-Apr-2018 Bryan Call

Updates to clang-tidy target

b82d0aaf11-Apr-2018 Masakazu Kitajo

Use #pragma once

711c4f4b09-Apr-2018 Bryan Call

Ran clang-tidy with readability-braces-around-statements

8d01164213-Mar-2018 dyrock

Added VConn args support
Changed TS_VCONN_PRE_ACCEPT to TS_VCONN_START
Added TS_VCONN_CLOSE hooks
Fixed docs for TS_VCONN_PRE_ACCEPT renaming to TS_VCONN_START and added docs for TSVconnA

Added VConn args support
Changed TS_VCONN_PRE_ACCEPT to TS_VCONN_START
Added TS_VCONN_CLOSE hooks
Fixed docs for TS_VCONN_PRE_ACCEPT renaming to TS_VCONN_START and added docs for TSVconnArgs
Added an example plugin vconn_args to test

Added vconn args man page

Make VCONN_PRE_ACCEPT deprecated but still compatible

show more ...

12d305b815-Mar-2018 Leif Hedstrom

Makes the global remap data structure ref-counted

This solves a number of issues, such as:

1. Remap plugins won't have to refcount itself
2. We can, I think, make rema

Makes the global remap data structure ref-counted

This solves a number of issues, such as:

1. Remap plugins won't have to refcount itself
2. We can, I think, make remap plugins reloadable

The code didn't turn out quite as I wanted; I had initially
intended to store the leased table in the t_state (::State)
object, but circular references here made that difficult
without major refactoring. Instead, I hoisted it up to the
HttpSM, and added a UrlRewrite* argument to a couple of
functions that needs it in addition to the t_state.

In addition, I eliminated the two macros around the ref-count
increment / decrement, seems unecessary in general, and only
used in a couple of places.

show more ...

7f295bf412-Mar-2018 Walt Karas

Remove leap-into-the-abyss casting from TSTransformCreate().

df144ab502-Mar-2018 James Peach

Fix default include path ordering.

Originally, prior to commit c7c9d0f7a, the configure phase
appended any include paths to `CPPFLAGS`. After c7c9d0f7a, they
were appended to `AM_CPP

Fix default include path ordering.

Originally, prior to commit c7c9d0f7a, the configure phase
appended any include paths to `CPPFLAGS`. After c7c9d0f7a, they
were appended to `AM_CPPFLAGS`. Since automake always specifies
`AM_CPPFLAGS` before `CPPFLAGS` when it generated the compilation
commands, this had the effect of always including system (or
library) headers before any in-tree headers. However, we actually
want the reverse (i.e. the original semantics) because in the
case of a name collision we will always want to resolve the
in-tree name.

This problem caused the build to break on macOS 10.13 (with a
case-insensitive filesystem) because the in-tree `Transform.h`
was being resolved to the Tcl `transform.h`. Hijinks ensued.

The fix is simple but slightly ugly. We keep using `AM_CPPFLAGS`
for global defines, but create a `TS_INCLUDES` build variable for
the include path. This lets us control the include ordering and
ensure that the autoconfigured include paths are included last.

show more ...

cf32494216-Feb-2018 Alan M. Carroll

BufferWriter: Documentation and some API cleanup.

96e0aca105-Feb-2018 scw00

correct negative event's timeout_at value

93565faa19-Jan-2018 Walt Karas

Make Continuation class constructor protected.

c7b2c64513-Jan-2018 Gancho Tenev

Reset water_mark in new_empty_MIOBuffer_internal

Moving water_mark reset one level up from MIOBuffer::alloc()
to new_MIOBuffer_internal() and adding it "symmetrically" to
new_empty_M

Reset water_mark in new_empty_MIOBuffer_internal

Moving water_mark reset one level up from MIOBuffer::alloc()
to new_MIOBuffer_internal() and adding it "symmetrically" to
new_empty_MIOBuffer_internal().
This will allow new_empty_MIOBuffer() to benefit from the
fix in commit db81103d0 as well.

show more ...

db81103d12-Jan-2018 Gancho Tenev

new_MIOBuffer: uninitialized water_mark fix.

While calling new_MIOBuffer() the water_mark is not initialized,
water_mark could end up with a very high (random) value to be used
durin

new_MIOBuffer: uninitialized water_mark fix.

While calling new_MIOBuffer() the water_mark is not initialized,
water_mark could end up with a very high (random) value to be used
during the network reads. This always results in overly aggressive
uninterrupted reading from the connection, severe buffer size increase
and CPU saturation of one or more ET_NET threads which renders the
traffic server unresponsive. Initializing the water_mark fixes the issue.

The problem can be easily reproduced by a series of very large TLS POST
requests in combination with a stalling origin server and periodic use
of FetchSM which internally allocates a buffer using new_MIOBuffer()
and sets water_mark to INT64_MAX.
While serving the POST requests a buffer previously used by FetchSM
would be reused and its water_mark would end up being equal to IN64_MAX
since the water mark is not initialized when calling new_MIOBuffer().
This always results in aggressive uninterrupted network reads, the
buffer size keep growing progressively and the constant block iteration
in IOBufferReader::read_avail() at some point would saturate the ET_NET
thread CPU.

This patch should fix issue #2144

Additionally fixed erroneous #ifdef TS_USE_PORT directive in
SSLNetVConnection::net_read_io().

show more ...

Revision tags: 7.1.2, 7.1.2-rc3, 7.1.2-rc2, 7.1.2-rc1, 7.1.2-rc0
3669b0ca15-Nov-2017 Alan M. Carroll

Fix compile error on Ubuntu.

2fd91a3a12-Nov-2017 Bryan Call

Cleanup of Continuation class initialization

ab7b372309-Nov-2017 Oknet Xu

Optimize: update ink_thread_create with thread-safe support

644b456d31-Oct-2017 Walt Karas

Remove MIOBufferWriter unit tests that throw exceptions.

ed592c1a21-Oct-2017 Oknet Xu

Implement zero-copy within UDPNetProcessorInternal::udp_read_from_net

Revision tags: 7.1.1
48c3e7b129-Aug-2017 Fei Deng

added tests for different thread configurations, thread information is verified using gdb

adding support for mac

use ps instead of gdb

use debug tag to print out thread inf

added tests for different thread configurations, thread information is verified using gdb

adding support for mac

use ps instead of gdb

use debug tag to print out thread information

clang-format check

show more ...

46a7fb8527-Oct-2017 Bryan Call

coverity 1376260: Uninitialized scalar field

dcb6f09c22-Oct-2017 Masaori Koshiba

CID 1226158: Uninitialized members

3ab9742519-Oct-2017 Fei Deng

event loop metrics collection

88431a4221-Sep-2017 Walt Karas

Add MIOBufferWriter class which writes to an MIOBuffer instance.

39992bb329-Aug-2017 Fei Deng

event loop changes

removed unnecessary defines

moved to class initializer

removed metrics collection

extracted externel queue processing code into new method

event loop changes

removed unnecessary defines

moved to class initializer

removed metrics collection

extracted externel queue processing code into new method

use std::min instead of just min

bug fix

remove unnecessary arg

show more ...

65b9b30403-Oct-2017 Phil Sorber

Remove volatile.

e638e7dd20-Sep-2017 Phil Sorber

Use std::min and std::max globally

12345678910>>...18