On 06/07/2017 03:11 PM, Emilio G. Cota wrote:
On Wed, Jun 07, 2017 at 08:55:30 -0700, Richard Henderson wrote: (snip)+#elif defined(__APPLE__) \ + || defined(__FreeBSD__) || defined(__FreeBSD_kernel__) +# include <sys/sysctl.h> +# if defined(__APPLE__) +# define SYSCTL_CACHELINE_NAME "hw.cachelinesize" +# else +# define SYSCTL_CACHELINE_NAME "machdep.cacheline_size" +# endif + +static void sys_cache_info(void) +{ + /* There's only a single sysctl for both I/D cache line sizes. */ + size_t len = sizeof(qemu_icache_linesize); + if (!sysctlbyname(SYSCTL_CACHELINE_NAME, &qemu_icache_linesize, + &len, NULL, 0)) { + qemu_dcache_linesize = qemu_icache_linesize; + } +}This doesn't work on the MacOS (Darwin 16.6.0) I have access to. If we do long size; // <-- type here matters! size_t len = sizeof(size); sysctlbyname(..., &size, &len, ...); then size == 64. However, if instead of 'long size' we have 'int size' (as in the patch), then 'size == 1' yet sysctlbyname still returns 0. (!) This is why I was using an intermediate long in my previous patch, although obviously a comment there would have been appropriate. FWIW, if we query what length sysctlbyname expects for this param (by calling sysctlbyname(NAME, NULL, &size, NULL, 0), we get len == 8; I presume on both 32 and 64-bit systems passing a long will work OK here.
Thanks, missed that. I also have to fix up the _WIN32 case, for which I missed a default: while rearranging things.
r~
