Mercurial > hg > nginx-quic
view src/os/unix/ngx_alloc.h @ 6440:248aa2757332
Fixed timeouts with threaded sendfile() and subrequests.
If a write event happens after sendfile() but before we've got the
sendfile results in the main thread, this write event will be ignored.
And if no more events will happen, the connection will hang.
Removing the events works in the simple cases, but not always, as
in some cases events are added back by an unrelated code. E.g.,
the upstream module adds write event in the ngx_http_upstream_init()
to track client aborts.
Fix is to use wev->complete instead. It is now set to 0 before
a sendfile() task is posted, and it is set to 1 once a write event
happens. If on completion of the sendfile() task wev->complete is 1,
we know that an event happened while we were executing sendfile(), and
the socket is still ready for writing even if sendfile() did not sent
all the data or returned EAGAIN.
author | Maxim Dounin <mdounin@mdounin.ru> |
---|---|
date | Fri, 18 Mar 2016 05:04:45 +0300 |
parents | d620f497c50f |
children |
line wrap: on
line source
/* * Copyright (C) Igor Sysoev * Copyright (C) Nginx, Inc. */ #ifndef _NGX_ALLOC_H_INCLUDED_ #define _NGX_ALLOC_H_INCLUDED_ #include <ngx_config.h> #include <ngx_core.h> void *ngx_alloc(size_t size, ngx_log_t *log); void *ngx_calloc(size_t size, ngx_log_t *log); #define ngx_free free /* * Linux has memalign() or posix_memalign() * Solaris has memalign() * FreeBSD 7.0 has posix_memalign(), besides, early version's malloc() * aligns allocations bigger than page size at the page boundary */ #if (NGX_HAVE_POSIX_MEMALIGN || NGX_HAVE_MEMALIGN) void *ngx_memalign(size_t alignment, size_t size, ngx_log_t *log); #else #define ngx_memalign(alignment, size, log) ngx_alloc(size, log) #endif extern ngx_uint_t ngx_pagesize; extern ngx_uint_t ngx_pagesize_shift; extern ngx_uint_t ngx_cacheline_size; #endif /* _NGX_ALLOC_H_INCLUDED_ */