On Mon, Jul 16, 2012 at 11:23 AM, Dunrong Huang <riegama...@gmail.com> wrote:
> 2012/7/16 Jing Huang <jing.huang....@gmail.com>:
>> This patch makes do_setsockopt() support SOL_RAW ICMP_FILTER socket option.
>>
>> Signed-off-by: Jing Huang <jing.huang....@gmail.com>
>> ---
>> linux-user/syscall.c | 20 ++++++++++++++++++++
>> 1 files changed, 20 insertions(+), 0 deletions(-)
>>
>> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
>> index 28c8ba5..fc8690d 100644
>> --- a/linux-user/syscall.c
>> +++ b/linux-user/syscall.c
>> @@ -60,6 +60,7 @@ int __clone2(int (*fn)(void *), void *child_stack_base,
>> #include <netinet/ip.h>
>> #include <netinet/tcp.h>
>> #include <linux/wireless.h>
>> +#include <linux/icmp.h>
>> #include "qemu-common.h"
>> #ifdef TARGET_GPROF
>> #include <sys/gmon.h>
>> @@ -1442,6 +1443,25 @@ static abi_long do_setsockopt(int sockfd, int level,
>> int optname,
>> goto unimplemented;
>> }
>> break;
>> + case SOL_RAW:
>> + switch (optname) {
>> + case ICMP_FILTER:
>> + /*struct icmp_filter takes an u32 value*/
>> + optname = ICMP_FILTER;
> Needless assignment statements. optname already has the value ICMP_FILTER
do_setsockopt() lacks level=SOL_RAW and optname=ICMP_FILTER.
The ping in linux-user guest invokes:
setsockopt(icmp_sock, SOL_RAW, ICMP_FILTER, (char *)&filt,
sizeof(filt)) in main() function.
>> + if (optlen < sizeof(uint32_t)) {
>> + return -TARGET_EINVAL;
>> + }
>> +
>> + if (get_user_u32(val, optval_addr)) {
>> + return -TARGET_EFAULT;
>> + }
>> + ret = get_errno(setsockopt(sockfd, level, optname,
>> + (char *)&val, sizeof(val)));
> Dont need casting &val to char *.
I refer to the definition of setsockopt() in linux-kernel/net/tipc/socket.c
>> + break;
>> + default:
>> + goto unimplemented;
>> + }
>> + break;
>> case TARGET_SOL_SOCKET:
>> switch (optname) {
>> /* Options with 'int' argument. */
>> --
>> 1.7.8.6
>>
>>
>
>
>
> --
> Best Regards,
>
> Dunrong Huang
