view src/core/ngx_cpuinfo.c @ 6874:7cc2d3a96ea3

Fixed trailer construction with limit on FreeBSD and macOS. The ngx_chain_coalesce_file() function may produce more bytes to send then requested in the limit passed, as it aligns the last file position to send to memory page boundary. As a result, (limit - send) may become negative. This resulted in big positive number when converted to size_t while calling ngx_output_chain_to_iovec(). Another part of the problem is in ngx_chain_coalesce_file(): it changes cl to the next chain link even if the current buffer is only partially sent due to limit. Therefore, if a file buffer was not expected to be fully sent due to limit, and was followed by a memory buffer, nginx called sendfile() with a part of the file buffer, and the memory buffer in trailer. If there were enough room in the socket buffer, this resulted in a part of the file buffer being skipped, and corresponding part of the memory buffer sent instead. The bug was introduced in 8e903522c17a (1.7.8). Configurations affected are ones using limits, that is, limit_rate and/or sendfile_max_chunk, and memory buffers after file ones (may happen when using subrequests or with proxying with disk buffering). Fix is to explicitly check if (send < limit) before constructing trailer with ngx_output_chain_to_iovec(). Additionally, ngx_chain_coalesce_file() was modified to preserve unfinished file buffers in cl.
author Maxim Dounin <mdounin@mdounin.ru>
date Fri, 20 Jan 2017 21:12:48 +0300
parents d620f497c50f
children
line wrap: on
line source


/*
 * Copyright (C) Igor Sysoev
 * Copyright (C) Nginx, Inc.
 */


#include <ngx_config.h>
#include <ngx_core.h>


#if (( __i386__ || __amd64__ ) && ( __GNUC__ || __INTEL_COMPILER ))


static ngx_inline void ngx_cpuid(uint32_t i, uint32_t *buf);


#if ( __i386__ )

static ngx_inline void
ngx_cpuid(uint32_t i, uint32_t *buf)
{

    /*
     * we could not use %ebx as output parameter if gcc builds PIC,
     * and we could not save %ebx on stack, because %esp is used,
     * when the -fomit-frame-pointer optimization is specified.
     */

    __asm__ (

    "    mov    %%ebx, %%esi;  "

    "    cpuid;                "
    "    mov    %%eax, (%1);   "
    "    mov    %%ebx, 4(%1);  "
    "    mov    %%edx, 8(%1);  "
    "    mov    %%ecx, 12(%1); "

    "    mov    %%esi, %%ebx;  "

    : : "a" (i), "D" (buf) : "ecx", "edx", "esi", "memory" );
}


#else /* __amd64__ */


static ngx_inline void
ngx_cpuid(uint32_t i, uint32_t *buf)
{
    uint32_t  eax, ebx, ecx, edx;

    __asm__ (

        "cpuid"

    : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx) : "a" (i) );

    buf[0] = eax;
    buf[1] = ebx;
    buf[2] = edx;
    buf[3] = ecx;
}


#endif


/* auto detect the L2 cache line size of modern and widespread CPUs */

void
ngx_cpuinfo(void)
{
    u_char    *vendor;
    uint32_t   vbuf[5], cpu[4], model;

    vbuf[0] = 0;
    vbuf[1] = 0;
    vbuf[2] = 0;
    vbuf[3] = 0;
    vbuf[4] = 0;

    ngx_cpuid(0, vbuf);

    vendor = (u_char *) &vbuf[1];

    if (vbuf[0] == 0) {
        return;
    }

    ngx_cpuid(1, cpu);

    if (ngx_strcmp(vendor, "GenuineIntel") == 0) {

        switch ((cpu[0] & 0xf00) >> 8) {

        /* Pentium */
        case 5:
            ngx_cacheline_size = 32;
            break;

        /* Pentium Pro, II, III */
        case 6:
            ngx_cacheline_size = 32;

            model = ((cpu[0] & 0xf0000) >> 8) | (cpu[0] & 0xf0);

            if (model >= 0xd0) {
                /* Intel Core, Core 2, Atom */
                ngx_cacheline_size = 64;
            }

            break;

        /*
         * Pentium 4, although its cache line size is 64 bytes,
         * it prefetches up to two cache lines during memory read
         */
        case 15:
            ngx_cacheline_size = 128;
            break;
        }

    } else if (ngx_strcmp(vendor, "AuthenticAMD") == 0) {
        ngx_cacheline_size = 64;
    }
}

#else


void
ngx_cpuinfo(void)
{
}


#endif