Mercurial > hg > nginx-vendor-0-8
view src/core/ngx_spinlock.c @ 276:c5c2b2883984 NGINX_0_5_8
nginx 0.5.8
*) Bugfix: a segmentation fault might occur if
"client_body_in_file_only on" was used and a request body was small.
*) Bugfix: a segmentation fault occurred if
"client_body_in_file_only on" and "proxy_pass_request_body off" or
"fastcgi_pass_request_body off" directives were used, and nginx
switched to a next upstream.
*) Bugfix: if the "proxy_buffering off" directive was used and a client
connection was non-active, then the connection was closed after send
timeout; bug appeared in 0.4.7.
*) Bugfix: if the "epoll" method was used and a client closed a
connection prematurely, then nginx closed the connection after a
send timeout only.
*) Bugfix: the "[alert] zero size buf" error when FastCGI server was
used.
*) Bugfixes in the "limit_zone" directive.
author | Igor Sysoev <http://sysoev.ru> |
---|---|
date | Fri, 19 Jan 2007 00:00:00 +0300 |
parents | 73e8476f9142 |
children |
line wrap: on
line source
/* * Copyright (C) Igor Sysoev */ #include <ngx_config.h> #include <ngx_core.h> void ngx_spinlock(ngx_atomic_t *lock, ngx_atomic_int_t value, ngx_uint_t spin) { #if (NGX_HAVE_ATOMIC_OPS) ngx_uint_t i, n; for ( ;; ) { if (*lock == 0 && ngx_atomic_cmp_set(lock, 0, value)) { return; } if (ngx_ncpu > 1) { for (n = 1; n < spin; n <<= 1) { for (i = 0; i < n; i++) { ngx_cpu_pause(); } if (*lock == 0 && ngx_atomic_cmp_set(lock, 0, value)) { return; } } } ngx_sched_yield(); } #else #if (NGX_THREADS) #error ngx_spinlock() or ngx_atomic_cmp_set() are not defined ! #endif #endif }