Re: [Qemu-devel] [PATCH v4] linux-user: Add sockopts for IPv6 ping and IPv6 traceroute

[Laurent Vivier]

Le 18/02/2017 à 09:31, Helge Deller a écrit :
> Add the neccessary sockopts for ping and traceroute on IPv6.
> 
> This fixes the following qemu warnings with IPv6:
> Unsupported ancillary data: 0/2
> Unsupported ancillary data: 0/11
> Unsupported ancillary data: 41/25
> Unsupported setsockopt level=0 optname=12 
> Unsupported setsockopt level=41 optname=16
> Unsupported setsockopt level=41 optname=25
> Unsupported setsockopt level=41 optname=50
> Unsupported setsockopt level=41 optname=51
> Unsupported setsockopt level=41 optname=8
> Unsupported setsockopt level=58 optname=1
> 
> Tested with hppa-linux-user (big-endian) on x86_64 (little-endian).
> 
> Signed-off-by: Helge Deller <address@hidden>
> 
> ---
> Changes to v3:
> - Fix up checkpatch fallouts
> 
> Changes to v2: (all suggested by Laurent Vivier)
> - Drop goto statements and replaced by real code
> - New function host_to_target_sockaddr_in6()
> - Fix IPV6_PKTINFO which uses in6_pktinfo instead of uint32_t
> - Move one IPV6_CHECKSUM from SOL_ICMPV6 to SOL_IPV6
> - Fix ICMPV6_FILTER to use icmpv6_filter
> 
> Changes to v1:
> - Added IPV6_PKTINFO sockopt as reported by Philippe Mathieu-Daudé
> 
> 
> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
> index f569f82..6bd3ea7 100644
> --- a/linux-user/syscall.c
> +++ b/linux-user/syscall.c
> @@ -57,6 +57,8 @@ int __clone2(int (*fn)(void *), void *child_stack_base,
>  #include <netinet/tcp.h>
>  #include <linux/wireless.h>
>  #include <linux/icmp.h>
> +#include <linux/icmpv6.h>
> +#include <linux/errqueue.h>
>  #include "qemu-common.h"
>  #ifdef CONFIG_TIMERFD
>  #include <sys/timerfd.h>
> @@ -1640,6 +1642,34 @@ static inline abi_long 
> host_to_target_sockaddr(abi_ulong target_addr,
>      return 0;
>  }
>  
> +static inline abi_long host_to_target_sockaddr_in6(abi_ulong target_addr,
> +                                               struct sockaddr_in6 *addr,
> +                                               socklen_t len)
> +{
> +    struct target_sockaddr_in6 *target_saddr;
> +
> +    if (len == 0) {
> +        return 0;
> +    }
> +
> +    target_saddr = lock_user(VERIFY_WRITE, target_addr, len, 0);
> +    if (!target_saddr) {
> +        return -TARGET_EFAULT;
> +    }
> +    memcpy(target_saddr, addr, len);
> +    if (len >= offsetof(struct target_sockaddr_in6, sin6_family) +
> +        sizeof(target_saddr->sin6_family)) {
> +        target_saddr->sin6_family = tswap16(addr->sin6_family);
> +    }
> +    if (len >= offsetof(struct target_sockaddr_in6, sin6_scope_id) +
> +        sizeof(target_saddr->sin6_scope_id)) {
> +        target_saddr->sin6_scope_id = tswap16(addr->sin6_scope_id);
> +    }
> +    unlock_user(target_saddr, target_addr, len);
> +
> +    return 0;
> +}


I think you can add directly this code in host_to_target_sockaddr() and
check if sa_family is AF_INET6 (and check length) and then swap
sin6_scope_id.

>  static inline abi_long target_to_host_cmsg(struct msghdr *msgh,
>                                             struct target_msghdr *target_msgh)
>  {
> @@ -1839,6 +1869,85 @@ static inline abi_long host_to_target_cmsg(struct 
> target_msghdr *target_msgh,
>              }
>              break;
>  
> +        case SOL_IP:
> +            switch (cmsg->cmsg_type) {
> +            case IP_TTL:
> +            {
> +                uint32_t *v = (uint32_t *)data;
> +                uint32_t *t_int = (uint32_t *)target_data;
> +                if (tgt_len != CMSG_LEN(0)) {

This checking looks strange, where does it come from?

You compare the target structure length (including data) with the host
structure head length.

> +                    goto unimplemented;
> +                }
> +
> +                __put_user(*v, t_int);
> +                break;
> +            }
> +            case IP_RECVERR:
> +            {
> +                struct errhdr_t {
> +                   struct sock_extended_err ee;
> +                   struct sockaddr_in offender;
> +                };
> +                struct errhdr_t *errh = (struct errhdr_t *)data;
> +                struct errhdr_t *target_errh =
> +                    (struct errhdr_t *)target_data;
> +
> +                __put_user(errh->ee.ee_errno, &target_errh->ee.ee_errno);
> +                __put_user(errh->ee.ee_origin, &target_errh->ee.ee_origin);
> +                __put_user(errh->ee.ee_type,  &target_errh->ee.ee_type);
> +                __put_user(errh->ee.ee_code, &target_errh->ee.ee_code);
> +                __put_user(errh->ee.ee_pad, &target_errh->ee.ee_pad);
> +                __put_user(errh->ee.ee_info, &target_errh->ee.ee_info);
> +                __put_user(errh->ee.ee_data, &target_errh->ee.ee_data);
> +                host_to_target_sockaddr((unsigned long) 
> &target_errh->offender,
> +                    (void *) &errh->offender, sizeof(errh->offender));
> +                break;
> +            }
> +            default:
> +                goto unimplemented;
> +            }
> +            break;
> +
> +        case SOL_IPV6:
> +            switch (cmsg->cmsg_type) {
> +            case IPV6_HOPLIMIT:
> +            {
> +                uint32_t *v = (uint32_t *)data;
> +                uint32_t *t_int = (uint32_t *)target_data;
> +                if (tgt_len != CMSG_LEN(0)) {
> +                    goto unimplemented;

ditto

> +                }
> +
> +                __put_user(*v, t_int);
> +                break;
> +            }
> +            case IPV6_RECVERR:
> +            {
> +                struct errhdr6_t {
> +                   struct sock_extended_err ee;
> +                   struct sockaddr_in6 offender;
> +                };
> +                struct errhdr6_t *errh = (struct errhdr6_t *)data;
> +                struct errhdr6_t *target_errh =
> +                    (struct errhdr6_t *)target_data;
> +
> +                __put_user(errh->ee.ee_errno, &target_errh->ee.ee_errno);
> +                __put_user(errh->ee.ee_origin, &target_errh->ee.ee_origin);
> +                __put_user(errh->ee.ee_type,  &target_errh->ee.ee_type);
> +                __put_user(errh->ee.ee_code, &target_errh->ee.ee_code);
> +                __put_user(errh->ee.ee_pad, &target_errh->ee.ee_pad);
> +                __put_user(errh->ee.ee_info, &target_errh->ee.ee_info);
> +                __put_user(errh->ee.ee_data, &target_errh->ee.ee_data);
> +                host_to_target_sockaddr_in6(
> +                    (unsigned long) &target_errh->offender,
> +                    &errh->offender, sizeof(errh->offender));
> +                break;
> +            }
> +            default:
> +                goto unimplemented;
> +            }
> +            break;
> +
>          default:
>          unimplemented:
>              gemu_log("Unsupported ancillary data: %d/%d\n",
> @@ -2768,6 +2877,7 @@ static abi_long do_setsockopt(int sockfd, int level, 
> int optname,
>          case IP_PKTINFO:
>          case IP_MTU_DISCOVER:
>          case IP_RECVERR:
> +        case IP_RECVTTL:
>          case IP_RECVTOS:
>  #ifdef IP_FREEBIND
>          case IP_FREEBIND:
> @@ -2817,6 +2927,11 @@ static abi_long do_setsockopt(int sockfd, int level, 
> int optname,
>          case IPV6_MTU:
>          case IPV6_V6ONLY:
>          case IPV6_RECVPKTINFO:
> +        case IPV6_UNICAST_HOPS:
> +        case IPV6_RECVERR:
> +        case IPV6_RECVHOPLIMIT:
> +        case IPV6_2292HOPLIMIT:
> +        case IPV6_CHECKSUM:
>              val = 0;
>              if (optlen < sizeof(uint32_t)) {
>                  return -TARGET_EINVAL;
> @@ -2827,6 +2942,50 @@ static abi_long do_setsockopt(int sockfd, int level, 
> int optname,
>              ret = get_errno(setsockopt(sockfd, level, optname,
>                                         &val, sizeof(val)));
>              break;
> +        case IPV6_PKTINFO:
> +        {
> +            struct in6_pktinfo pki;
> +
> +            if (optlen < sizeof(pki)) {
> +                return -TARGET_EINVAL;
> +            }
> +
> +            if (copy_from_user(&pki, optval_addr, sizeof(pki))) {
> +                return -TARGET_EFAULT;
> +            }
> +
> +            pki.ipi6_ifindex = tswap32(pki.ipi6_ifindex);
> +
> +            ret = get_errno(setsockopt(sockfd, level, optname,
> +                                       &pki, sizeof(pki)));
> +            break;
> +        }
> +        default:
> +            goto unimplemented;
> +        }
> +        break;
> +    case SOL_ICMPV6:
> +        switch (optname) {
> +        case ICMPV6_FILTER:
> +        {
> +            struct icmp6_filter icmp6f;
> +
> +            if (optlen < sizeof(icmp6f)) {
> +                return -TARGET_EINVAL;

kernel doesn't return EINVAL, it trunks the length instead.


Laurent