* [PATCH 0/6] fcntl, sockopt, and ioctl options
@ 2020-07-23 0:19 Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls Shu-Chun Weng
` (6 more replies)
0 siblings, 7 replies; 11+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:19 UTC (permalink / raw)
To: qemu-devel; +Cc: Shu-Chun Weng, laurent
Hi Laurent,
This is a series of 6 patches in 4 groups, putting into a single thread for
easier tracking.
[PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls
An incidental follow up on
https://lists.nongnu.org/archive/html/qemu-devel/2019-09/msg01925.html
[PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt
Updated https://lists.nongnu.org/archive/html/qemu-devel/2019-09/msg01317.html
to consistently add them in get/setsockopt
[PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW
[PATCH 4/6] linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING
Updated https://lists.nongnu.org/archive/html/qemu-devel/2019-09/msg01319.html
to only use TARGET_SO_*_OLD/NEW
[PATCH 5/6] thunk: supports flexible arrays
[PATCH 6/6] linux-user: Add support for SIOCETHTOOL ioctl
Updated https://lists.nongnu.org/archive/html/qemu-devel/2019-08/msg05090.html
Shu-Chun Weng (6):
linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls
linux-user: add missing UDP and IPv6 get/setsockopt options
linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW
linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING
thunk: supports flexible arrays
linux-user: Add support for SIOCETHTOOL ioctl
include/exec/user/thunk.h | 20 +
linux-user/Makefile.objs | 3 +-
linux-user/alpha/sockbits.h | 21 +-
linux-user/ethtool.c | 819 +++++++++++++++++++++++++
linux-user/ethtool.h | 19 +
linux-user/ethtool_entries.h | 107 ++++
linux-user/fd-trans.h | 41 +-
linux-user/generic/sockbits.h | 17 +-
linux-user/hppa/sockbits.h | 20 +-
linux-user/ioctls.h | 2 +
linux-user/mips/sockbits.h | 16 +-
linux-user/qemu.h | 1 +
linux-user/sparc/sockbits.h | 21 +-
linux-user/strace.c | 19 +-
linux-user/syscall.c | 233 ++++++-
linux-user/syscall_defs.h | 26 +-
linux-user/syscall_types.h | 277 +++++++++
tests/tcg/multiarch/ethtool.c | 417 +++++++++++++
tests/tcg/multiarch/socket_timestamp.c | 542 ++++++++++++++++
thunk.c | 151 ++++-
20 files changed, 2706 insertions(+), 66 deletions(-)
create mode 100644 linux-user/ethtool.c
create mode 100644 linux-user/ethtool.h
create mode 100644 linux-user/ethtool_entries.h
create mode 100644 tests/tcg/multiarch/ethtool.c
create mode 100644 tests/tcg/multiarch/socket_timestamp.c
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
@ 2020-07-23 0:19 ` Shu-Chun Weng
2020-08-07 11:44 ` Laurent Vivier
2020-07-23 0:19 ` [PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt options Shu-Chun Weng
` (5 subsequent siblings)
6 siblings, 1 reply; 11+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:19 UTC (permalink / raw)
To: qemu-devel; +Cc: Shu-Chun Weng, laurent
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
linux-user/syscall.c | 10 ++++++++++
linux-user/syscall_defs.h | 14 ++++++++------
2 files changed, 18 insertions(+), 6 deletions(-)
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 1211e759c2..f97337b0b4 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -6312,6 +6312,14 @@ static int target_to_host_fcntl_cmd(int cmd)
case TARGET_F_GETPIPE_SZ:
ret = F_GETPIPE_SZ;
break;
+#endif
+#ifdef F_ADD_SEALS
+ case TARGET_F_ADD_SEALS:
+ ret = F_ADD_SEALS;
+ break;
+ case TARGET_F_GET_SEALS:
+ ret = F_GET_SEALS;
+ break;
#endif
default:
ret = -TARGET_EINVAL;
@@ -6598,6 +6606,8 @@ static abi_long do_fcntl(int fd, int cmd, abi_ulong arg)
case TARGET_F_GETLEASE:
case TARGET_F_SETPIPE_SZ:
case TARGET_F_GETPIPE_SZ:
+ case TARGET_F_ADD_SEALS:
+ case TARGET_F_GET_SEALS:
ret = get_errno(safe_fcntl(fd, host_cmd, arg));
break;
diff --git a/linux-user/syscall_defs.h b/linux-user/syscall_defs.h
index 3c261cff0e..70df1a94fb 100644
--- a/linux-user/syscall_defs.h
+++ b/linux-user/syscall_defs.h
@@ -2292,12 +2292,14 @@ struct target_statfs64 {
#endif
#define TARGET_F_LINUX_SPECIFIC_BASE 1024
-#define TARGET_F_SETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 0)
-#define TARGET_F_GETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 1)
-#define TARGET_F_DUPFD_CLOEXEC (TARGET_F_LINUX_SPECIFIC_BASE + 6)
-#define TARGET_F_SETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 7)
-#define TARGET_F_GETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 8)
-#define TARGET_F_NOTIFY (TARGET_F_LINUX_SPECIFIC_BASE+2)
+#define TARGET_F_SETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 0)
+#define TARGET_F_GETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 1)
+#define TARGET_F_DUPFD_CLOEXEC (TARGET_F_LINUX_SPECIFIC_BASE + 6)
+#define TARGET_F_NOTIFY (TARGET_F_LINUX_SPECIFIC_BASE + 2)
+#define TARGET_F_SETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 7)
+#define TARGET_F_GETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 8)
+#define TARGET_F_ADD_SEALS (TARGET_F_LINUX_SPECIFIC_BASE + 9)
+#define TARGET_F_GET_SEALS (TARGET_F_LINUX_SPECIFIC_BASE + 10)
#include "target_fcntl.h"
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt options
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls Shu-Chun Weng
@ 2020-07-23 0:19 ` Shu-Chun Weng
2020-08-07 11:54 ` Laurent Vivier
2020-07-23 0:19 ` [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW Shu-Chun Weng
` (4 subsequent siblings)
6 siblings, 1 reply; 11+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:19 UTC (permalink / raw)
To: qemu-devel; +Cc: Shu-Chun Weng, laurent
UDP: SOL_UDP manipulate options at UDP level. All six options currently
defined in linux source include/uapi/linux/udp.h take integer values.
IPv6: IPV6_ADDR_PREFERENCES (RFC5014: Source address selection) was not
supported.
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
linux-user/syscall.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index f97337b0b4..a53db446d4 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -51,8 +51,10 @@
#include <sys/sysinfo.h>
#include <sys/signalfd.h>
//#include <sys/user.h>
+#include <netinet/in.h>
#include <netinet/ip.h>
#include <netinet/tcp.h>
+#include <netinet/udp.h>
#include <linux/wireless.h>
#include <linux/icmp.h>
#include <linux/icmpv6.h>
@@ -1945,7 +1947,8 @@ static abi_long do_setsockopt(int sockfd, int level, int optname,
switch(level) {
case SOL_TCP:
- /* TCP options all take an 'int' value. */
+ case SOL_UDP:
+ /* TCP and UDP options all take an 'int' value. */
if (optlen < sizeof(uint32_t))
return -TARGET_EINVAL;
@@ -2031,6 +2034,7 @@ static abi_long do_setsockopt(int sockfd, int level, int optname,
case IPV6_RECVDSTOPTS:
case IPV6_2292DSTOPTS:
case IPV6_TCLASS:
+ case IPV6_ADDR_PREFERENCES:
#ifdef IPV6_RECVPATHMTU
case IPV6_RECVPATHMTU:
#endif
@@ -2593,7 +2597,8 @@ get_timeout:
}
break;
case SOL_TCP:
- /* TCP options all take an 'int' value. */
+ case SOL_UDP:
+ /* TCP and UDP options all take an 'int' value. */
int_case:
if (get_user_u32(len, optlen))
return -TARGET_EFAULT;
@@ -2684,6 +2689,7 @@ get_timeout:
case IPV6_RECVDSTOPTS:
case IPV6_2292DSTOPTS:
case IPV6_TCLASS:
+ case IPV6_ADDR_PREFERENCES:
#ifdef IPV6_RECVPATHMTU
case IPV6_RECVPATHMTU:
#endif
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt options Shu-Chun Weng
@ 2020-07-23 0:19 ` Shu-Chun Weng
2020-08-07 12:07 ` Laurent Vivier
2020-07-23 0:19 ` [PATCH 4/6] linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING Shu-Chun Weng
` (3 subsequent siblings)
6 siblings, 1 reply; 11+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:19 UTC (permalink / raw)
To: qemu-devel; +Cc: Shu-Chun Weng, laurent, alex.bennee
Both guest options map to host SO_TIMESTAMP while keeping a bit in
fd_trans to remember if the guest expects the old or the new format.
Added a multiarch test to verify.
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
linux-user/alpha/sockbits.h | 8 +-
linux-user/fd-trans.h | 41 +++-
linux-user/generic/sockbits.h | 9 +-
linux-user/hppa/sockbits.h | 8 +-
linux-user/mips/sockbits.h | 8 +-
linux-user/sparc/sockbits.h | 8 +-
linux-user/strace.c | 7 +-
linux-user/syscall.c | 69 ++++--
tests/tcg/multiarch/socket_timestamp.c | 292 +++++++++++++++++++++++++
9 files changed, 419 insertions(+), 31 deletions(-)
create mode 100644 tests/tcg/multiarch/socket_timestamp.c
diff --git a/linux-user/alpha/sockbits.h b/linux-user/alpha/sockbits.h
index d54dc98c09..40f0644df0 100644
--- a/linux-user/alpha/sockbits.h
+++ b/linux-user/alpha/sockbits.h
@@ -48,8 +48,6 @@
#define TARGET_SO_DETACH_FILTER 27
#define TARGET_SO_PEERNAME 28
-#define TARGET_SO_TIMESTAMP 29
-#define TARGET_SCM_TIMESTAMP TARGET_SO_TIMESTAMP
#define TARGET_SO_PEERSEC 30
#define TARGET_SO_PASSSEC 34
@@ -75,6 +73,12 @@
/* Instruct lower device to use last 4-bytes of skb data as FCS */
#define TARGET_SO_NOFCS 43
+#define TARGET_SO_TIMESTAMP_OLD 29
+#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+
+#define TARGET_SO_TIMESTAMP_NEW 63
+#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+
/* TARGET_O_NONBLOCK clashes with the bits used for socket types. Therefore we
* have to define SOCK_NONBLOCK to a different value here.
*/
diff --git a/linux-user/fd-trans.h b/linux-user/fd-trans.h
index a3fcdaabc7..8ab650dfd2 100644
--- a/linux-user/fd-trans.h
+++ b/linux-user/fd-trans.h
@@ -22,6 +22,16 @@ typedef struct TargetFdTrans {
TargetFdDataFunc host_to_target_data;
TargetFdDataFunc target_to_host_data;
TargetFdAddrFunc target_to_host_addr;
+
+ /* If `true`, this struct is dynamically allocated and should be
+ * `g_free()`ed when unregistering.
+ */
+ bool free_when_unregister;
+
+ /* The socket's timestamp option (`SO_TIMESTAMP`, `SO_TIMESTAMPNS`, and
+ * `SO_TIMESTAMPING`) is using the `_NEW` version.
+ */
+ bool socket_timestamp_new;
} TargetFdTrans;
extern TargetFdTrans **target_fd_trans;
@@ -52,6 +62,14 @@ static inline TargetFdAddrFunc fd_trans_target_to_host_addr(int fd)
return NULL;
}
+static inline bool fd_trans_socket_timestamp_new(int fd)
+{
+ if (fd >= 0 && fd < target_fd_max && target_fd_trans[fd]) {
+ return target_fd_trans[fd]->socket_timestamp_new;
+ }
+ return false;
+}
+
static inline void fd_trans_register(int fd, TargetFdTrans *trans)
{
unsigned int oldmax;
@@ -70,6 +88,9 @@ static inline void fd_trans_register(int fd, TargetFdTrans *trans)
static inline void fd_trans_unregister(int fd)
{
if (fd >= 0 && fd < target_fd_max) {
+ if (target_fd_trans[fd] && target_fd_trans[fd]->free_when_unregister) {
+ g_free(target_fd_trans[fd]);
+ }
target_fd_trans[fd] = NULL;
}
}
@@ -78,8 +99,26 @@ static inline void fd_trans_dup(int oldfd, int newfd)
{
fd_trans_unregister(newfd);
if (oldfd < target_fd_max && target_fd_trans[oldfd]) {
- fd_trans_register(newfd, target_fd_trans[oldfd]);
+ TargetFdTrans *trans = target_fd_trans[oldfd];
+ if (trans->free_when_unregister) {
+ trans = g_new(TargetFdTrans, 1);
+ *trans = *target_fd_trans[oldfd];
+ }
+ fd_trans_register(newfd, trans);
+ }
+}
+
+static inline void fd_trans_mark_socket_timestamp_new(int fd, bool value)
+{
+ if (fd < 0) return;
+ if (fd >= target_fd_max || target_fd_trans[fd] == NULL) {
+ if (!value) return; /* default is false */
+
+ TargetFdTrans* trans = g_new0(TargetFdTrans, 1);
+ trans->free_when_unregister = true;
+ fd_trans_register(fd, trans);
}
+ target_fd_trans[fd]->socket_timestamp_new = value;
}
extern TargetFdTrans target_packet_trans;
diff --git a/linux-user/generic/sockbits.h b/linux-user/generic/sockbits.h
index e44733c601..532cf2d3dc 100644
--- a/linux-user/generic/sockbits.h
+++ b/linux-user/generic/sockbits.h
@@ -49,10 +49,15 @@
#define TARGET_SO_DETACH_FILTER 27
#define TARGET_SO_PEERNAME 28
-#define TARGET_SO_TIMESTAMP 29
-#define TARGET_SCM_TIMESTAMP TARGET_SO_TIMESTAMP
#define TARGET_SO_ACCEPTCONN 30
#define TARGET_SO_PEERSEC 31
+
+#define TARGET_SO_TIMESTAMP_OLD 29
+#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+
+#define TARGET_SO_TIMESTAMP_NEW 63
+#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+
#endif
diff --git a/linux-user/hppa/sockbits.h b/linux-user/hppa/sockbits.h
index 23f69a3293..284a47e74e 100644
--- a/linux-user/hppa/sockbits.h
+++ b/linux-user/hppa/sockbits.h
@@ -29,8 +29,6 @@
#define TARGET_SO_BSDCOMPAT 0x400e
#define TARGET_SO_PASSCRED 0x4010
#define TARGET_SO_PEERCRED 0x4011
-#define TARGET_SO_TIMESTAMP 0x4012
-#define TARGET_SCM_TIMESTAMP TARGET_SO_TIMESTAMP
#define TARGET_SO_TIMESTAMPNS 0x4013
#define TARGET_SCM_TIMESTAMPNS TARGET_SO_TIMESTAMPNS
@@ -67,6 +65,12 @@
#define TARGET_SO_CNX_ADVICE 0x402E
+#define TARGET_SO_TIMESTAMP_OLD 0x4012
+#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+
+#define TARGET_SO_TIMESTAMP_NEW 0x4038
+#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+
/* TARGET_O_NONBLOCK clashes with the bits used for socket types. Therefore we
* have to define SOCK_NONBLOCK to a different value here.
*/
diff --git a/linux-user/mips/sockbits.h b/linux-user/mips/sockbits.h
index 0f022cd598..b4c39d9588 100644
--- a/linux-user/mips/sockbits.h
+++ b/linux-user/mips/sockbits.h
@@ -61,14 +61,18 @@
#define TARGET_SO_DETACH_FILTER 27
#define TARGET_SO_PEERNAME 28
-#define TARGET_SO_TIMESTAMP 29
-#define SCM_TIMESTAMP SO_TIMESTAMP
#define TARGET_SO_PEERSEC 30
#define TARGET_SO_SNDBUFFORCE 31
#define TARGET_SO_RCVBUFFORCE 33
#define TARGET_SO_PASSSEC 34
+#define TARGET_SO_TIMESTAMP_OLD 29
+#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+
+#define TARGET_SO_TIMESTAMP_NEW 63
+#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+
/** sock_type - Socket types
*
* Please notice that for binary compat reasons MIPS has to
diff --git a/linux-user/sparc/sockbits.h b/linux-user/sparc/sockbits.h
index 0a822e3e1f..07440efd14 100644
--- a/linux-user/sparc/sockbits.h
+++ b/linux-user/sparc/sockbits.h
@@ -48,8 +48,6 @@
#define TARGET_SO_GET_FILTER TARGET_SO_ATTACH_FILTER
#define TARGET_SO_PEERNAME 0x001c
-#define TARGET_SO_TIMESTAMP 0x001d
-#define TARGET_SCM_TIMESTAMP TARGET_SO_TIMESTAMP
#define TARGET_SO_PEERSEC 0x001e
#define TARGET_SO_PASSSEC 0x001f
@@ -104,6 +102,12 @@
#define TARGET_SO_ZEROCOPY 0x003e
+#define TARGET_SO_TIMESTAMP_OLD 0x001d
+#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+
+#define TARGET_SO_TIMESTAMP_NEW 0x0046
+#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+
/* Security levels - as per NRL IPv6 - don't actually do anything */
#define TARGET_SO_SECURITY_AUTHENTICATION 0x5001
#define TARGET_SO_SECURITY_ENCRYPTION_TRANSPORT 0x5002
diff --git a/linux-user/strace.c b/linux-user/strace.c
index 13981341b3..369cfcd7bd 100644
--- a/linux-user/strace.c
+++ b/linux-user/strace.c
@@ -2213,8 +2213,11 @@ print_optint:
case TARGET_SO_PASSCRED:
qemu_log("SO_PASSCRED,");
goto print_optint;
- case TARGET_SO_TIMESTAMP:
- qemu_log("SO_TIMESTAMP,");
+ case TARGET_SO_TIMESTAMP_OLD:
+ qemu_log("SO_TIMESTAMP_OLD,");
+ goto print_optint;
+ case TARGET_SO_TIMESTAMP_NEW:
+ qemu_log("SO_TIMESTAMP_NEW,");
goto print_optint;
case TARGET_SO_RCVLOWAT:
qemu_log("SO_RCVLOWAT,");
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index a53db446d4..7f95ccda7f 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -1705,7 +1705,8 @@ static inline abi_long target_to_host_cmsg(struct msghdr *msgh,
}
static inline abi_long host_to_target_cmsg(struct target_msghdr *target_msgh,
- struct msghdr *msgh)
+ struct msghdr *msgh,
+ bool target_uses_timeval64)
{
struct cmsghdr *cmsg = CMSG_FIRSTHDR(msgh);
abi_long msg_controllen;
@@ -1754,8 +1755,14 @@ static inline abi_long host_to_target_cmsg(struct target_msghdr *target_msgh,
switch (cmsg->cmsg_level) {
case SOL_SOCKET:
switch (cmsg->cmsg_type) {
- case SO_TIMESTAMP:
- tgt_len = sizeof(struct target_timeval);
+ case SCM_TIMESTAMP:
+ if (target_uses_timeval64) {
+ tgt_len = sizeof(struct target__kernel_sock_timeval);
+ target_cmsg->cmsg_type = tswap32(TARGET_SCM_TIMESTAMP_NEW);
+ } else {
+ tgt_len = sizeof(struct target_timeval);
+ target_cmsg->cmsg_type = tswap32(TARGET_SCM_TIMESTAMP_OLD);
+ }
break;
default:
break;
@@ -1789,20 +1796,32 @@ static inline abi_long host_to_target_cmsg(struct target_msghdr *target_msgh,
}
break;
}
- case SO_TIMESTAMP:
+ case SCM_TIMESTAMP:
{
struct timeval *tv = (struct timeval *)data;
- struct target_timeval *target_tv =
- (struct target_timeval *)target_data;
-
- if (len != sizeof(struct timeval) ||
- tgt_len != sizeof(struct target_timeval)) {
+ if (len != sizeof(struct timeval)) {
goto unimplemented;
}
- /* copy struct timeval to target */
- __put_user(tv->tv_sec, &target_tv->tv_sec);
- __put_user(tv->tv_usec, &target_tv->tv_usec);
+ if (target_uses_timeval64) {
+ struct target__kernel_sock_timeval *target_tv =
+ (struct target__kernel_sock_timeval *)target_data;
+ if (tgt_len != sizeof(struct target__kernel_sock_timeval)) {
+ goto unimplemented;
+ }
+
+ __put_user(tv->tv_sec, &target_tv->tv_sec);
+ __put_user(tv->tv_usec, &target_tv->tv_usec);
+ } else {
+ struct target_timeval *target_tv =
+ (struct target_timeval *)target_data;
+ if (tgt_len != sizeof(struct target_timeval)) {
+ goto unimplemented;
+ }
+
+ __put_user(tv->tv_sec, &target_tv->tv_sec);
+ __put_user(tv->tv_usec, &target_tv->tv_usec);
+ }
break;
}
case SCM_CREDENTIALS:
@@ -1941,7 +1960,7 @@ static abi_long do_setsockopt(int sockfd, int level, int optname,
abi_ulong optval_addr, socklen_t optlen)
{
abi_long ret;
- int val;
+ int val, timestamp_format_is_new;
struct ip_mreqn *ip_mreq;
struct ip_mreq_source *ip_mreq_source;
@@ -2338,9 +2357,11 @@ set_timeout:
case TARGET_SO_PASSSEC:
optname = SO_PASSSEC;
break;
- case TARGET_SO_TIMESTAMP:
- optname = SO_TIMESTAMP;
- break;
+ case TARGET_SO_TIMESTAMP_OLD:
+ case TARGET_SO_TIMESTAMP_NEW:
+ timestamp_format_is_new = (optname == TARGET_SO_TIMESTAMP_NEW);
+ optname = SO_TIMESTAMP;
+ break;
case TARGET_SO_RCVLOWAT:
optname = SO_RCVLOWAT;
break;
@@ -2353,6 +2374,9 @@ set_timeout:
if (get_user_u32(val, optval_addr))
return -TARGET_EFAULT;
ret = get_errno(setsockopt(sockfd, SOL_SOCKET, optname, &val, sizeof(val)));
+ if (!is_error(ret) && optname == SO_TIMESTAMP) {
+ fd_trans_mark_socket_timestamp_new(sockfd, timestamp_format_is_new);
+ }
break;
#ifdef SOL_NETLINK
case SOL_NETLINK:
@@ -2403,6 +2427,7 @@ static abi_long do_getsockopt(int sockfd, int level, int optname,
abi_long ret;
int len, val;
socklen_t lv;
+ int timestamp_format_matches = 0;
switch(level) {
case TARGET_SOL_SOCKET:
@@ -2583,7 +2608,11 @@ get_timeout:
case TARGET_SO_PASSCRED:
optname = SO_PASSCRED;
goto int_case;
- case TARGET_SO_TIMESTAMP:
+ case TARGET_SO_TIMESTAMP_OLD:
+ case TARGET_SO_TIMESTAMP_NEW:
+ timestamp_format_matches =
+ (fd_trans_socket_timestamp_new(sockfd) ==
+ (optname == TARGET_SO_TIMESTAMP_NEW));
optname = SO_TIMESTAMP;
goto int_case;
case TARGET_SO_RCVLOWAT:
@@ -2611,6 +2640,9 @@ get_timeout:
if (optname == SO_TYPE) {
val = host_to_target_sock_type(val);
}
+ if (optname == SO_TIMESTAMP) {
+ val = val && timestamp_format_matches;
+ }
if (len > lv)
len = lv;
if (len == 4) {
@@ -3162,7 +3194,8 @@ static abi_long do_sendrecvmsg_locked(int fd, struct target_msghdr *msgp,
ret = fd_trans_host_to_target_data(fd)(msg.msg_iov->iov_base,
MIN(msg.msg_iov->iov_len, len));
} else {
- ret = host_to_target_cmsg(msgp, &msg);
+ ret = host_to_target_cmsg(msgp, &msg,
+ fd_trans_socket_timestamp_new(fd));
}
if (!is_error(ret)) {
msgp->msg_namelen = tswap32(msg.msg_namelen);
diff --git a/tests/tcg/multiarch/socket_timestamp.c b/tests/tcg/multiarch/socket_timestamp.c
new file mode 100644
index 0000000000..fd2833e5c8
--- /dev/null
+++ b/tests/tcg/multiarch/socket_timestamp.c
@@ -0,0 +1,292 @@
+#include <assert.h>
+#include <errno.h>
+#include <linux/types.h>
+#include <netinet/in.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <sys/time.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+
+#ifdef __kernel_old_timeval
+#define kernel_old_timeval __kernel_old_timeval
+#else
+struct kernel_old_timeval
+{
+ __kernel_long_t tv_sec;
+ __kernel_long_t tv_usec;
+};
+#endif
+
+struct kernel_sock_timeval
+{
+ int64_t tv_sec;
+ int64_t tv_usec;
+};
+
+int create_udp_socket(struct sockaddr_in *sockaddr)
+{
+ socklen_t sockaddr_len;
+ int sock = socket(AF_INET, SOCK_DGRAM, 0);
+ if (sock < 0) {
+ int err = errno;
+ fprintf(stderr, "Failed to create server socket: %s\n", strerror(err));
+ exit(err);
+ }
+
+ memset(sockaddr, 0, sizeof(*sockaddr));
+ sockaddr->sin_family = AF_INET;
+ sockaddr->sin_port = htons(0); /* let kernel select a port for us */
+ sockaddr->sin_addr.s_addr = htonl(INADDR_LOOPBACK);
+
+ if (bind(sock, (struct sockaddr *)sockaddr, sizeof(*sockaddr)) < 0) {
+ int err = errno;
+ fprintf(stderr, "Failed to bind server socket: %s\n", strerror(err));
+ exit(err);
+ }
+
+ sockaddr_len = sizeof(*sockaddr);
+ if (getsockname(sock, (struct sockaddr *)sockaddr, &sockaddr_len) < 0) {
+ int err = errno;
+ fprintf(stderr, "Failed to get socket name: %s\n", strerror(err));
+ exit(err);
+ }
+ return sock;
+}
+
+/* Checks that the timestamp in the message is not after the reception timestamp
+ * as well as the reception time is within 10 seconds of the message time.
+ */
+void check_timestamp_difference(const struct timeval *msg_tv,
+ const struct timeval *pkt_tv)
+{
+ if (pkt_tv->tv_sec < msg_tv->tv_sec ||
+ (pkt_tv->tv_sec == msg_tv->tv_sec && pkt_tv->tv_usec < msg_tv->tv_usec))
+ {
+ fprintf(stderr,
+ "Packet received before sent: %lld.%06lld < %lld.%06lld\n",
+ (long long)pkt_tv->tv_sec, (long long)pkt_tv->tv_usec,
+ (long long)msg_tv->tv_sec, (long long)msg_tv->tv_usec);
+ exit(-1);
+ }
+
+ if (pkt_tv->tv_sec > msg_tv->tv_sec + 10 ||
+ (pkt_tv->tv_sec == msg_tv->tv_sec + 10 &&
+ pkt_tv->tv_usec > msg_tv->tv_usec)) {
+ fprintf(stderr,
+ "Packet received more than 10 seconds after sent: "
+ "%lld.%06lld > %lld.%06lld + 10\n",
+ (long long)pkt_tv->tv_sec, (long long)pkt_tv->tv_usec,
+ (long long)msg_tv->tv_sec, (long long)msg_tv->tv_usec);
+ exit(-1);
+ }
+}
+
+void send_current_time(int sock, struct sockaddr_in server_sockaddr)
+{
+ struct timeval tv = {0, 0};
+ gettimeofday(&tv, NULL);
+ sendto(sock, &tv, sizeof(tv), 0, (struct sockaddr *)&server_sockaddr,
+ sizeof(server_sockaddr));
+}
+
+typedef void (*get_timeval_t)(const struct cmsghdr *cmsg, struct timeval *tv);
+
+
+void receive_packet(int sock, get_timeval_t get_timeval)
+{
+ struct msghdr msg = {0};
+
+ char iobuf[1024];
+ struct iovec iov;
+
+ union {
+ /* 128 bytes should cover all imaginable timeval/timespec*/
+ char cmsg_buf[CMSG_SPACE(128)];
+ struct cmsghdr align;
+ } u;
+ struct cmsghdr *cmsg;
+ struct timeval msg_tv, pkt_tv;
+
+ int res;
+
+ iov.iov_base = iobuf;
+ iov.iov_len = sizeof(iobuf);
+
+ msg.msg_iov = &iov;
+ msg.msg_iovlen = 1;
+ msg.msg_control = (caddr_t)u.cmsg_buf;
+ msg.msg_controllen = sizeof(u.cmsg_buf);
+
+ if ((res = recvmsg(sock, &msg, 0)) < 0) {
+ int err = errno;
+ fprintf(stderr, "Failed to receive packet: %s\n", strerror(err));
+ exit(-err);
+ }
+
+ assert(res == sizeof(struct timeval));
+ assert(iov.iov_base == iobuf);
+ memcpy(&msg_tv, iov.iov_base, sizeof(msg_tv));
+ printf("Message timestamp: %lld.%06lld\n",
+ (long long)msg_tv.tv_sec, (long long)msg_tv.tv_usec);
+
+ cmsg = CMSG_FIRSTHDR(&msg);
+ assert(cmsg);
+ (*get_timeval)(cmsg, &pkt_tv);
+ printf("Packet timestamp: %lld.%06lld\n",
+ (long long)pkt_tv.tv_sec, (long long)pkt_tv.tv_usec);
+
+ check_timestamp_difference(&msg_tv, &pkt_tv);
+}
+
+void get_timeval_from_so_timestamp(const struct cmsghdr *cmsg,
+ struct timeval *tv)
+{
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SCM_TIMESTAMP);
+ assert(cmsg->cmsg_len >= CMSG_LEN(sizeof(struct timeval)));
+ memcpy(tv, CMSG_DATA(cmsg), sizeof(*tv));
+}
+
+#ifdef SO_TIMESTAMP_OLD
+void get_timeval_from_so_timestamp_old(const struct cmsghdr *cmsg,
+ struct timeval *tv)
+{
+ struct kernel_old_timeval old_tv;
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SO_TIMESTAMP_OLD);
+ assert(cmsg->cmsg_len >= CMSG_LEN(sizeof(old_tv)));
+
+ memcpy(&old_tv, CMSG_DATA(cmsg), sizeof(old_tv));
+ tv->tv_sec = old_tv.tv_sec;
+ tv->tv_usec = old_tv.tv_usec;
+}
+
+#ifdef SO_TIMESTAMP_NEW
+void get_timeval_from_so_timestamp_new(const struct cmsghdr *cmsg,
+ struct timeval *tv)
+{
+ struct kernel_sock_timeval sock_tv;
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SO_TIMESTAMP_NEW);
+ assert(cmsg->cmsg_len >= CMSG_LEN(sizeof(sock_tv)));
+
+ memcpy(&sock_tv, CMSG_DATA(cmsg), sizeof(sock_tv));
+ tv->tv_sec = sock_tv.tv_sec;
+ tv->tv_usec = sock_tv.tv_usec;
+}
+#endif /* defined(SO_TIMESTAMP_NEW) */
+#endif /* defined(SO_TIMESTAMP_OLD) */
+
+void set_socket_option(int sock, int sockopt, int on)
+{
+ socklen_t len;
+ int val = on;
+ if (setsockopt(sock, SOL_SOCKET, sockopt, &val, sizeof(val)) < 0) {
+ int err = errno;
+ fprintf(stderr, "Failed to setsockopt %d (%s): %s\n",
+ sockopt, on ? "on" : "off", strerror(err));
+ exit(err);
+ }
+
+ len = sizeof(val);
+ val = -1;
+ if (getsockopt(sock, SOL_SOCKET, sockopt, &val, &len) < 0) {
+ int err = errno;
+ fprintf(stderr, "Failed to getsockopt (%d): %s\n", sock, strerror(err));
+ exit(err);
+ }
+ assert(len == sizeof(val));
+ assert(val == on);
+}
+
+int main(int argc, char **argv)
+{
+ int parent_sock, child_sock;
+ struct sockaddr_in parent_sockaddr, child_sockaddr;
+ int pid;
+ struct timeval tv = {0, 0};
+ gettimeofday(&tv, NULL);
+
+ parent_sock = create_udp_socket(&parent_sockaddr);
+ child_sock = create_udp_socket(&child_sockaddr);
+
+ printf("Parent sock bound to port %d\nChild sock bound to port %d\n",
+ parent_sockaddr.sin_port, child_sockaddr.sin_port);
+
+ if ((pid = fork()) == -2) {
+ fprintf(stderr, "SKIPPED. Failed to fork: %s\n", strerror(errno));
+ } else if (pid == 0) {
+ close(child_sock);
+
+ /* Test 1: SO_TIMESTAMP */
+ send_current_time(parent_sock, child_sockaddr);
+
+ if (tv.tv_sec > 0x7fffff00) {
+ /* Too close to y2038 problem, old system may not work. */
+ close(parent_sock);
+ return 0;
+ }
+
+#ifdef SO_TIMESTAMP_OLD
+ if (SO_TIMESTAMP_OLD != SO_TIMESTAMP) {
+ /* Test 2a: SO_TIMESTAMP_OLD */
+ set_socket_option(parent_sock, SO_TIMESTAMP_OLD, 1);
+ receive_packet(parent_sock, &get_timeval_from_so_timestamp_old);
+ set_socket_option(parent_sock, SO_TIMESTAMP_OLD, 0);
+ }
+#ifdef SO_TIMESTAMP_NEW
+ else {
+ /* Test 2b: SO_TIMESTAMP_NEW */
+ set_socket_option(parent_sock, SO_TIMESTAMP_NEW, 1);
+ receive_packet(parent_sock, &get_timeval_from_so_timestamp_new);
+ set_socket_option(parent_sock, SO_TIMESTAMP_NEW, 0);
+ }
+#endif /* defined(SO_TIMESTAMP_NEW) */
+#endif /* defined(SO_TIMESTAMP_OLD) */
+
+ close(parent_sock);
+ } else {
+ int child_status;
+ close(parent_sock);
+
+ /* Test 1: SO_TIMESTAMP */
+ set_socket_option(child_sock, SO_TIMESTAMP, 1);
+ receive_packet(child_sock, &get_timeval_from_so_timestamp);
+ set_socket_option(child_sock, SO_TIMESTAMP, 0);
+
+ if (tv.tv_sec > 0x7fffff00) {
+ /* Too close to y2038 problem, old system may not work. */
+ close(child_sock);
+ return 0;
+ }
+
+#ifdef SO_TIMESTAMP_OLD
+ if (SO_TIMESTAMP_OLD != SO_TIMESTAMP) {
+ /* Test 2a: SO_TIMESTAMP_OLD */
+ send_current_time(child_sock, parent_sockaddr);
+ }
+#ifdef SO_TIMESTAMP_NEW
+ else {
+ /* Test 2b: SO_TIMESTAMP_NEW */
+ send_current_time(child_sock, parent_sockaddr);
+ }
+#endif /* defined(SO_TIMESTAMP_NEW) */
+#endif /* defined(SO_TIMESTAMP_OLD) */
+
+ close(child_sock);
+
+ if (waitpid(pid, &child_status, 0) < 0) {
+ int err = errno;
+ fprintf(stderr, "Final wait() failed: %s\n", strerror(err));
+ return err;
+ }
+ return child_status;
+ }
+ return 0;
+}
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 4/6] linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
` (2 preceding siblings ...)
2020-07-23 0:19 ` [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW Shu-Chun Weng
@ 2020-07-23 0:19 ` Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 5/6] thunk: supports flexible arrays Shu-Chun Weng
` (2 subsequent siblings)
6 siblings, 0 replies; 11+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:19 UTC (permalink / raw)
To: qemu-devel; +Cc: Shu-Chun Weng, laurent, alex.bennee
This change supports SO_TIMESTAMPNS_OLD/NEW and SO_TIMESTAMPING_OLD/NEW
for setsocketopt() with SOL_SOCKET. Based on the SO_TIMESTAMP_OLD/NEW
framework. The three pairs share the same flag `SOCK_TSTAMP_NEW` in
linux kernel for deciding if the old or the new format is used.
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
linux-user/alpha/sockbits.h | 13 +-
linux-user/generic/sockbits.h | 8 +
linux-user/hppa/sockbits.h | 12 +-
linux-user/mips/sockbits.h | 8 +
linux-user/sparc/sockbits.h | 13 +-
linux-user/strace.c | 12 +
linux-user/syscall.c | 119 ++++++-
tests/tcg/multiarch/socket_timestamp.c | 458 +++++++++++++++++++------
8 files changed, 521 insertions(+), 122 deletions(-)
diff --git a/linux-user/alpha/sockbits.h b/linux-user/alpha/sockbits.h
index 40f0644df0..c2c88f432b 100644
--- a/linux-user/alpha/sockbits.h
+++ b/linux-user/alpha/sockbits.h
@@ -51,8 +51,6 @@
#define TARGET_SO_PEERSEC 30
#define TARGET_SO_PASSSEC 34
-#define TARGET_SO_TIMESTAMPNS 35
-#define TARGET_SCM_TIMESTAMPNS TARGET_SO_TIMESTAMPNS
/* Security levels - as per NRL IPv6 - don't actually do anything */
#define TARGET_SO_SECURITY_AUTHENTICATION 19
@@ -61,9 +59,6 @@
#define TARGET_SO_MARK 36
-#define TARGET_SO_TIMESTAMPING 37
-#define TARGET_SCM_TIMESTAMPING TARGET_SO_TIMESTAMPING
-
#define TARGET_SO_RXQ_OVFL 40
#define TARGET_SO_WIFI_STATUS 41
@@ -75,9 +70,17 @@
#define TARGET_SO_TIMESTAMP_OLD 29
#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+#define TARGET_SO_TIMESTAMPNS_OLD 35
+#define TARGET_SCM_TIMESTAMPNS_OLD TARGET_SO_TIMESTAMPNS_OLD
+#define TARGET_SO_TIMESTAMPING_OLD 37
+#define TARGET_SCM_TIMESTAMPING_OLD TARGET_SO_TIMESTAMPING_OLD
#define TARGET_SO_TIMESTAMP_NEW 63
#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+#define TARGET_SO_TIMESTAMPNS_NEW 64
+#define TARGET_SCM_TIMESTAMPNS_NEW TARGET_SO_TIMESTAMPNS_NEW
+#define TARGET_SO_TIMESTAMPING_NEW 65
+#define TARGET_SCM_TIMESTAMPING_NEW TARGET_SO_TIMESTAMPING_NEW
/* TARGET_O_NONBLOCK clashes with the bits used for socket types. Therefore we
* have to define SOCK_NONBLOCK to a different value here.
diff --git a/linux-user/generic/sockbits.h b/linux-user/generic/sockbits.h
index 532cf2d3dc..a0496d8751 100644
--- a/linux-user/generic/sockbits.h
+++ b/linux-user/generic/sockbits.h
@@ -56,8 +56,16 @@
#define TARGET_SO_TIMESTAMP_OLD 29
#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+#define TARGET_SO_TIMESTAMPNS_OLD 35
+#define TARGET_SCM_TIMESTAMPNS_OLD TARGET_SO_TIMESTAMPNS_OLD
+#define TARGET_SO_TIMESTAMPING_OLD 37
+#define TARGET_SCM_TIMESTAMPING_OLD TARGET_SO_TIMESTAMPING_OLD
#define TARGET_SO_TIMESTAMP_NEW 63
#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+#define TARGET_SO_TIMESTAMPNS_NEW 64
+#define TARGET_SCM_TIMESTAMPNS_NEW TARGET_SO_TIMESTAMPNS_NEW
+#define TARGET_SO_TIMESTAMPING_NEW 65
+#define TARGET_SCM_TIMESTAMPING_NEW TARGET_SO_TIMESTAMPING_NEW
#endif
diff --git a/linux-user/hppa/sockbits.h b/linux-user/hppa/sockbits.h
index 284a47e74e..d7e9aa340d 100644
--- a/linux-user/hppa/sockbits.h
+++ b/linux-user/hppa/sockbits.h
@@ -29,8 +29,6 @@
#define TARGET_SO_BSDCOMPAT 0x400e
#define TARGET_SO_PASSCRED 0x4010
#define TARGET_SO_PEERCRED 0x4011
-#define TARGET_SO_TIMESTAMPNS 0x4013
-#define TARGET_SCM_TIMESTAMPNS TARGET_SO_TIMESTAMPNS
#define TARGET_SO_SECURITY_AUTHENTICATION 0x4016
#define TARGET_SO_SECURITY_ENCRYPTION_TRANSPORT 0x4017
@@ -44,8 +42,6 @@
#define TARGET_SO_PEERSEC 0x401d
#define TARGET_SO_PASSSEC 0x401e
#define TARGET_SO_MARK 0x401f
-#define TARGET_SO_TIMESTAMPING 0x4020
-#define TARGET_SCM_TIMESTAMPING TARGET_SO_TIMESTAMPING
#define TARGET_SO_RXQ_OVFL 0x4021
#define TARGET_SO_WIFI_STATUS 0x4022
#define TARGET_SCM_WIFI_STATUS TARGET_SO_WIFI_STATUS
@@ -67,9 +63,17 @@
#define TARGET_SO_TIMESTAMP_OLD 0x4012
#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+#define TARGET_SO_TIMESTAMPNS_OLD 0x4013
+#define TARGET_SCM_TIMESTAMPNS_OLD TARGET_SO_TIMESTAMPNS_OLD
+#define TARGET_SO_TIMESTAMPING_OLD 0x4020
+#define TARGET_SCM_TIMESTAMPING_OLD TARGET_SO_TIMESTAMPING_OLD
#define TARGET_SO_TIMESTAMP_NEW 0x4038
#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+#define TARGET_SO_TIMESTAMPNS_NEW 0x4039
+#define TARGET_SCM_TIMESTAMPNS_NEW TARGET_SO_TIMESTAMPNS_NEW
+#define TARGET_SO_TIMESTAMPING_NEW 0x403A
+#define TARGET_SCM_TIMESTAMPING_NEW TARGET_SO_TIMESTAMPING_NEW
/* TARGET_O_NONBLOCK clashes with the bits used for socket types. Therefore we
* have to define SOCK_NONBLOCK to a different value here.
diff --git a/linux-user/mips/sockbits.h b/linux-user/mips/sockbits.h
index b4c39d9588..49524e23ac 100644
--- a/linux-user/mips/sockbits.h
+++ b/linux-user/mips/sockbits.h
@@ -69,9 +69,17 @@
#define TARGET_SO_TIMESTAMP_OLD 29
#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+#define TARGET_SO_TIMESTAMPNS_OLD 35
+#define TARGET_SCM_TIMESTAMPNS_OLD TARGET_SO_TIMESTAMPNS_OLD
+#define TARGET_SO_TIMESTAMPING_OLD 37
+#define TARGET_SCM_TIMESTAMPING_OLD TARGET_SO_TIMESTAMPING_OLD
#define TARGET_SO_TIMESTAMP_NEW 63
#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+#define TARGET_SO_TIMESTAMPNS_NEW 64
+#define TARGET_SCM_TIMESTAMPNS_NEW TARGET_SO_TIMESTAMPNS_NEW
+#define TARGET_SO_TIMESTAMPING_NEW 65
+#define TARGET_SCM_TIMESTAMPING_NEW TARGET_SO_TIMESTAMPING_NEW
/** sock_type - Socket types
*
diff --git a/linux-user/sparc/sockbits.h b/linux-user/sparc/sockbits.h
index 07440efd14..c5fade3ad1 100644
--- a/linux-user/sparc/sockbits.h
+++ b/linux-user/sparc/sockbits.h
@@ -51,14 +51,9 @@
#define TARGET_SO_PEERSEC 0x001e
#define TARGET_SO_PASSSEC 0x001f
-#define TARGET_SO_TIMESTAMPNS 0x0021
-#define TARGET_SCM_TIMESTAMPNS TARGET_SO_TIMESTAMPNS
#define TARGET_SO_MARK 0x0022
-#define TARGET_SO_TIMESTAMPING 0x0023
-#define TARGET_SCM_TIMESTAMPING TARGET_SO_TIMESTAMPING
-
#define TARGET_SO_RXQ_OVFL 0x0024
#define TARGET_SO_WIFI_STATUS 0x0025
@@ -104,9 +99,17 @@
#define TARGET_SO_TIMESTAMP_OLD 0x001d
#define TARGET_SCM_TIMESTAMP_OLD TARGET_SO_TIMESTAMP_OLD
+#define TARGET_SO_TIMESTAMPNS_OLD 0x0021
+#define TARGET_SCM_TIMESTAMPNS_OLD TARGET_SO_TIMESTAMPNS_OLD
+#define TARGET_SO_TIMESTAMPING_OLD 0x0023
+#define TARGET_SCM_TIMESTAMPING_OLD TARGET_SO_TIMESTAMPING_OLD
#define TARGET_SO_TIMESTAMP_NEW 0x0046
#define TARGET_SCM_TIMESTAMP_NEW TARGET_SO_TIMESTAMP_NEW
+#define TARGET_SO_TIMESTAMPNS_NEW 0x0042
+#define TARGET_SCM_TIMESTAMPNS_NEW TARGET_SO_TIMESTAMPNS_NEW
+#define TARGET_SO_TIMESTAMPING_NEW 0x0043
+#define TARGET_SCM_TIMESTAMPING_NEW TARGET_SO_TIMESTAMPING_NEW
/* Security levels - as per NRL IPv6 - don't actually do anything */
#define TARGET_SO_SECURITY_AUTHENTICATION 0x5001
diff --git a/linux-user/strace.c b/linux-user/strace.c
index 369cfcd7bd..603723d6c8 100644
--- a/linux-user/strace.c
+++ b/linux-user/strace.c
@@ -2216,9 +2216,21 @@ print_optint:
case TARGET_SO_TIMESTAMP_OLD:
qemu_log("SO_TIMESTAMP_OLD,");
goto print_optint;
+ case TARGET_SO_TIMESTAMPNS_OLD:
+ qemu_log("SO_TIMESTAMPNS_OLD,");
+ goto print_optint;
+ case TARGET_SO_TIMESTAMPING_OLD:
+ qemu_log("SO_TIMESTAMPING_OLD,");
+ goto print_optint;
case TARGET_SO_TIMESTAMP_NEW:
qemu_log("SO_TIMESTAMP_NEW,");
goto print_optint;
+ case TARGET_SO_TIMESTAMPNS_NEW:
+ qemu_log("SO_TIMESTAMPNS_NEW,");
+ goto print_optint;
+ case TARGET_SO_TIMESTAMPING_NEW:
+ qemu_log("SO_TIMESTAMPING_NEW,");
+ goto print_optint;
case TARGET_SO_RCVLOWAT:
qemu_log("SO_RCVLOWAT,");
goto print_optint;
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 7f95ccda7f..483f8ce48f 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -1764,6 +1764,28 @@ static inline abi_long host_to_target_cmsg(struct target_msghdr *target_msgh,
target_cmsg->cmsg_type = tswap32(TARGET_SCM_TIMESTAMP_OLD);
}
break;
+ case SCM_TIMESTAMPNS:
+ if (target_uses_timeval64) {
+ tgt_len = sizeof(struct target__kernel_timespec);
+ target_cmsg->cmsg_type =
+ tswap32(TARGET_SCM_TIMESTAMPNS_NEW);
+ } else {
+ tgt_len = sizeof(struct target_timespec);
+ target_cmsg->cmsg_type =
+ tswap32(TARGET_SCM_TIMESTAMPNS_OLD);
+ }
+ break;
+ case SCM_TIMESTAMPING:
+ if (target_uses_timeval64) {
+ tgt_len = sizeof(struct target__kernel_timespec[3]);
+ target_cmsg->cmsg_type =
+ tswap32(TARGET_SCM_TIMESTAMPING_NEW);
+ } else {
+ tgt_len = sizeof(struct target_timespec[3]);
+ target_cmsg->cmsg_type =
+ tswap32(TARGET_SCM_TIMESTAMPING_OLD);
+ }
+ break;
default:
break;
}
@@ -1824,6 +1846,67 @@ static inline abi_long host_to_target_cmsg(struct target_msghdr *target_msgh,
}
break;
}
+ case SCM_TIMESTAMPNS:
+ {
+ struct timespec *ts = (struct timespec *)data;
+ if (len != sizeof(struct timespec)) {
+ goto unimplemented;
+ }
+
+ if (target_uses_timeval64) {
+ struct target__kernel_timespec *target_ts =
+ (struct target__kernel_timespec *)target_data;
+ if (tgt_len != sizeof(struct target__kernel_timespec)) {
+ goto unimplemented;
+ }
+
+ __put_user(ts->tv_sec, &target_ts->tv_sec);
+ __put_user(ts->tv_nsec, &target_ts->tv_nsec);
+ } else {
+ struct target_timespec *target_ts =
+ (struct target_timespec *)target_data;
+ if (tgt_len != sizeof(struct target_timespec)) {
+ goto unimplemented;
+ }
+
+ __put_user(ts->tv_sec, &target_ts->tv_sec);
+ __put_user(ts->tv_nsec, &target_ts->tv_nsec);
+ }
+ break;
+ }
+ case SCM_TIMESTAMPING:
+ {
+ int i;
+ struct timespec *ts = (struct timespec *)data;
+ if (len != sizeof(struct timespec[3])) {
+ goto unimplemented;
+ }
+
+ if (target_uses_timeval64) {
+ struct target__kernel_timespec *target_ts =
+ (struct target__kernel_timespec *)target_data;
+ if (tgt_len != sizeof(struct target__kernel_timespec[3])) {
+ goto unimplemented;
+ }
+
+ for (i = 0; i < 3; ++i) {
+ __put_user(ts[i].tv_sec, &target_ts[i].tv_sec);
+ __put_user(ts[i].tv_nsec, &target_ts[i].tv_nsec);
+ }
+ } else {
+ struct target_timespec *target_ts =
+ (struct target_timespec *)target_data;
+ if (tgt_len != sizeof(struct target_timespec[3])) {
+ goto unimplemented;
+ }
+
+ for (i = 0; i < 3; ++i) {
+ __put_user(ts[i].tv_sec, &target_ts[i].tv_sec);
+ __put_user(ts[i].tv_nsec, &target_ts[i].tv_nsec);
+ }
+ }
+ break;
+ }
case SCM_CREDENTIALS:
{
struct ucred *cred = (struct ucred *)data;
@@ -2362,6 +2445,18 @@ set_timeout:
timestamp_format_is_new = (optname == TARGET_SO_TIMESTAMP_NEW);
optname = SO_TIMESTAMP;
break;
+ case TARGET_SO_TIMESTAMPNS_OLD:
+ case TARGET_SO_TIMESTAMPNS_NEW:
+ timestamp_format_is_new =
+ (optname == TARGET_SO_TIMESTAMPNS_NEW);
+ optname = SO_TIMESTAMPNS;
+ break;
+ case TARGET_SO_TIMESTAMPING_OLD:
+ case TARGET_SO_TIMESTAMPING_NEW:
+ timestamp_format_is_new =
+ (optname == TARGET_SO_TIMESTAMPING_NEW);
+ optname = SO_TIMESTAMPING;
+ break;
case TARGET_SO_RCVLOWAT:
optname = SO_RCVLOWAT;
break;
@@ -2374,7 +2469,9 @@ set_timeout:
if (get_user_u32(val, optval_addr))
return -TARGET_EFAULT;
ret = get_errno(setsockopt(sockfd, SOL_SOCKET, optname, &val, sizeof(val)));
- if (!is_error(ret) && optname == SO_TIMESTAMP) {
+ if (!is_error(ret) &&
+ (optname == SO_TIMESTAMP || optname == SO_TIMESTAMPNS ||
+ optname == SO_TIMESTAMPING)) {
fd_trans_mark_socket_timestamp_new(sockfd, timestamp_format_is_new);
}
break;
@@ -2615,6 +2712,20 @@ get_timeout:
(optname == TARGET_SO_TIMESTAMP_NEW));
optname = SO_TIMESTAMP;
goto int_case;
+ case TARGET_SO_TIMESTAMPNS_OLD:
+ case TARGET_SO_TIMESTAMPNS_NEW:
+ timestamp_format_matches =
+ (fd_trans_socket_timestamp_new(sockfd) ==
+ (optname == TARGET_SO_TIMESTAMPNS_NEW));
+ optname = SO_TIMESTAMPNS;
+ goto int_case;
+ case TARGET_SO_TIMESTAMPING_OLD:
+ case TARGET_SO_TIMESTAMPING_NEW:
+ timestamp_format_matches =
+ (fd_trans_socket_timestamp_new(sockfd) ==
+ (optname == TARGET_SO_TIMESTAMPING_NEW));
+ optname = SO_TIMESTAMPING;
+ goto int_case;
case TARGET_SO_RCVLOWAT:
optname = SO_RCVLOWAT;
goto int_case;
@@ -2639,9 +2750,9 @@ get_timeout:
return ret;
if (optname == SO_TYPE) {
val = host_to_target_sock_type(val);
- }
- if (optname == SO_TIMESTAMP) {
- val = val && timestamp_format_matches;
+ } else if ((optname == SO_TIMESTAMP || optname == SO_TIMESTAMPNS ||
+ optname == SO_TIMESTAMPING) && !timestamp_format_matches) {
+ val = 0;
}
if (len > lv)
len = lv;
diff --git a/tests/tcg/multiarch/socket_timestamp.c b/tests/tcg/multiarch/socket_timestamp.c
index fd2833e5c8..fe6059226b 100644
--- a/tests/tcg/multiarch/socket_timestamp.c
+++ b/tests/tcg/multiarch/socket_timestamp.c
@@ -1,5 +1,6 @@
#include <assert.h>
#include <errno.h>
+#include <linux/net_tstamp.h>
#include <linux/types.h>
#include <netinet/in.h>
#include <stdint.h>
@@ -11,6 +12,7 @@
#include <sys/time.h>
#include <sys/types.h>
#include <sys/wait.h>
+#include <time.h>
#include <unistd.h>
#ifdef __kernel_old_timeval
@@ -29,6 +31,38 @@ struct kernel_sock_timeval
int64_t tv_usec;
};
+struct kernel_old_timespec
+{
+ __kernel_long_t tv_sec;
+ long tv_nsec;
+};
+
+struct kernel_timespec
+{
+ int64_t tv_sec;
+ long long tv_nsec;
+};
+
+struct scm_timestamping
+{
+ struct timespec ts[3];
+};
+
+struct scm_old_timestamping
+{
+ struct kernel_old_timespec ts[3];
+};
+
+struct scm_timestamping64
+{
+ struct kernel_timespec ts[3];
+};
+
+const int so_timestamping_flags =
+ SOF_TIMESTAMPING_RX_HARDWARE |
+ SOF_TIMESTAMPING_RX_SOFTWARE |
+ SOF_TIMESTAMPING_SOFTWARE;
+
int create_udp_socket(struct sockaddr_in *sockaddr)
{
socklen_t sockaddr_len;
@@ -62,43 +96,47 @@ int create_udp_socket(struct sockaddr_in *sockaddr)
/* Checks that the timestamp in the message is not after the reception timestamp
* as well as the reception time is within 10 seconds of the message time.
*/
-void check_timestamp_difference(const struct timeval *msg_tv,
- const struct timeval *pkt_tv)
+void check_timestamp_difference(const struct timespec *msg_ts,
+ const struct timespec *pkt_ts)
{
- if (pkt_tv->tv_sec < msg_tv->tv_sec ||
- (pkt_tv->tv_sec == msg_tv->tv_sec && pkt_tv->tv_usec < msg_tv->tv_usec))
+ if (pkt_ts->tv_sec < msg_ts->tv_sec ||
+ (pkt_ts->tv_sec == msg_ts->tv_sec && pkt_ts->tv_nsec < msg_ts->tv_nsec))
{
fprintf(stderr,
- "Packet received before sent: %lld.%06lld < %lld.%06lld\n",
- (long long)pkt_tv->tv_sec, (long long)pkt_tv->tv_usec,
- (long long)msg_tv->tv_sec, (long long)msg_tv->tv_usec);
+ "Packet received before sent: %lld.%06lld < %lld.%09lld\n",
+ (long long)pkt_ts->tv_sec, (long long)pkt_ts->tv_nsec,
+ (long long)msg_ts->tv_sec, (long long)msg_ts->tv_nsec);
exit(-1);
}
- if (pkt_tv->tv_sec > msg_tv->tv_sec + 10 ||
- (pkt_tv->tv_sec == msg_tv->tv_sec + 10 &&
- pkt_tv->tv_usec > msg_tv->tv_usec)) {
+ if (pkt_ts->tv_sec > msg_ts->tv_sec + 10 ||
+ (pkt_ts->tv_sec == msg_ts->tv_sec + 10 &&
+ pkt_ts->tv_nsec > msg_ts->tv_nsec)) {
fprintf(stderr,
"Packet received more than 10 seconds after sent: "
- "%lld.%06lld > %lld.%06lld + 10\n",
- (long long)pkt_tv->tv_sec, (long long)pkt_tv->tv_usec,
- (long long)msg_tv->tv_sec, (long long)msg_tv->tv_usec);
+ "%lld.%06lld > %lld.%09lld + 10\n",
+ (long long)pkt_ts->tv_sec, (long long)pkt_ts->tv_nsec,
+ (long long)msg_ts->tv_sec, (long long)msg_ts->tv_nsec);
exit(-1);
}
}
void send_current_time(int sock, struct sockaddr_in server_sockaddr)
{
- struct timeval tv = {0, 0};
- gettimeofday(&tv, NULL);
- sendto(sock, &tv, sizeof(tv), 0, (struct sockaddr *)&server_sockaddr,
+ struct timespec ts = {0, 0};
+ clock_gettime(CLOCK_REALTIME, &ts);
+#ifdef MSG_CONFIRM
+ const int flags = MSG_CONFIRM;
+#else
+ const int flags = 0;
+#endif
+ sendto(sock, &ts, sizeof(ts), flags, (struct sockaddr *)&server_sockaddr,
sizeof(server_sockaddr));
}
-typedef void (*get_timeval_t)(const struct cmsghdr *cmsg, struct timeval *tv);
-
+typedef void (*get_timespec_t)(const struct cmsghdr *cmsg, struct timespec *tv);
-void receive_packet(int sock, get_timeval_t get_timeval)
+void receive_packet(int sock, get_timespec_t get_timespec)
{
struct msghdr msg = {0};
@@ -106,12 +144,14 @@ void receive_packet(int sock, get_timeval_t get_timeval)
struct iovec iov;
union {
- /* 128 bytes should cover all imaginable timeval/timespec*/
+ /* 128 bytes are enough for all existing
+ * timeval/timespec/scm_timestamping structures.
+ */
char cmsg_buf[CMSG_SPACE(128)];
struct cmsghdr align;
} u;
struct cmsghdr *cmsg;
- struct timeval msg_tv, pkt_tv;
+ struct timespec msg_ts, pkt_ts;
int res;
@@ -126,160 +166,370 @@ void receive_packet(int sock, get_timeval_t get_timeval)
if ((res = recvmsg(sock, &msg, 0)) < 0) {
int err = errno;
fprintf(stderr, "Failed to receive packet: %s\n", strerror(err));
- exit(-err);
+ exit(err);
}
assert(res == sizeof(struct timeval));
assert(iov.iov_base == iobuf);
- memcpy(&msg_tv, iov.iov_base, sizeof(msg_tv));
- printf("Message timestamp: %lld.%06lld\n",
- (long long)msg_tv.tv_sec, (long long)msg_tv.tv_usec);
+ memcpy(&msg_ts, iov.iov_base, sizeof(msg_ts));
+ printf("Message timestamp: %lld.%09lld\n",
+ (long long)msg_ts.tv_sec, (long long)msg_ts.tv_nsec);
cmsg = CMSG_FIRSTHDR(&msg);
assert(cmsg);
- (*get_timeval)(cmsg, &pkt_tv);
- printf("Packet timestamp: %lld.%06lld\n",
- (long long)pkt_tv.tv_sec, (long long)pkt_tv.tv_usec);
+ (*get_timespec)(cmsg, &pkt_ts);
+ printf("Packet timestamp: %lld.%09lld\n",
+ (long long)pkt_ts.tv_sec, (long long)pkt_ts.tv_nsec);
- check_timestamp_difference(&msg_tv, &pkt_tv);
+ check_timestamp_difference(&msg_ts, &pkt_ts);
}
-void get_timeval_from_so_timestamp(const struct cmsghdr *cmsg,
- struct timeval *tv)
+void get_timespec_from_so_timestamp(const struct cmsghdr *cmsg,
+ struct timespec *ts)
{
+ struct timeval tv;
assert(cmsg->cmsg_level == SOL_SOCKET);
assert(cmsg->cmsg_type == SCM_TIMESTAMP);
- assert(cmsg->cmsg_len >= CMSG_LEN(sizeof(struct timeval)));
- memcpy(tv, CMSG_DATA(cmsg), sizeof(*tv));
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(tv)));
+
+ memcpy(&tv, CMSG_DATA(cmsg), sizeof(tv));
+ ts->tv_sec = tv.tv_sec;
+ ts->tv_nsec = tv.tv_usec * 1000LL;
}
#ifdef SO_TIMESTAMP_OLD
-void get_timeval_from_so_timestamp_old(const struct cmsghdr *cmsg,
- struct timeval *tv)
+void get_timespec_from_so_timestamp_old(const struct cmsghdr *cmsg,
+ struct timespec *ts)
{
struct kernel_old_timeval old_tv;
assert(cmsg->cmsg_level == SOL_SOCKET);
assert(cmsg->cmsg_type == SO_TIMESTAMP_OLD);
- assert(cmsg->cmsg_len >= CMSG_LEN(sizeof(old_tv)));
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(old_tv)));
memcpy(&old_tv, CMSG_DATA(cmsg), sizeof(old_tv));
- tv->tv_sec = old_tv.tv_sec;
- tv->tv_usec = old_tv.tv_usec;
+ ts->tv_sec = old_tv.tv_sec;
+ ts->tv_nsec = old_tv.tv_usec * 1000LL;
}
#ifdef SO_TIMESTAMP_NEW
-void get_timeval_from_so_timestamp_new(const struct cmsghdr *cmsg,
- struct timeval *tv)
+void get_timespec_from_so_timestamp_new(const struct cmsghdr *cmsg,
+ struct timespec *ts)
{
struct kernel_sock_timeval sock_tv;
assert(cmsg->cmsg_level == SOL_SOCKET);
assert(cmsg->cmsg_type == SO_TIMESTAMP_NEW);
- assert(cmsg->cmsg_len >= CMSG_LEN(sizeof(sock_tv)));
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(sock_tv)));
memcpy(&sock_tv, CMSG_DATA(cmsg), sizeof(sock_tv));
- tv->tv_sec = sock_tv.tv_sec;
- tv->tv_usec = sock_tv.tv_usec;
+ ts->tv_sec = sock_tv.tv_sec;
+ ts->tv_nsec = sock_tv.tv_usec * 1000LL;
}
#endif /* defined(SO_TIMESTAMP_NEW) */
#endif /* defined(SO_TIMESTAMP_OLD) */
-void set_socket_option(int sock, int sockopt, int on)
+void get_timespec_from_so_timestampns(const struct cmsghdr *cmsg,
+ struct timespec *ts)
+{
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SCM_TIMESTAMPNS);
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(*ts)));
+
+ memcpy(ts, CMSG_DATA(cmsg), sizeof(*ts));
+}
+
+#ifdef SO_TIMESTAMPNS_OLD
+void get_timespec_from_so_timestampns_old(const struct cmsghdr *cmsg,
+ struct timespec *ts)
+{
+ struct kernel_old_timespec old_ts;
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SO_TIMESTAMPNS_OLD);
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(old_ts)));
+
+ memcpy(&old_ts, CMSG_DATA(cmsg), sizeof(old_ts));
+ ts->tv_sec = old_ts.tv_sec;
+ ts->tv_nsec = old_ts.tv_nsec;
+}
+
+#ifdef SO_TIMESTAMPNS_NEW
+void get_timespec_from_so_timestampns_new(const struct cmsghdr *cmsg,
+ struct timespec *ts)
+{
+ struct kernel_timespec sock_ts;
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SO_TIMESTAMPNS_NEW);
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(sock_ts)));
+
+ memcpy(&sock_ts, CMSG_DATA(cmsg), sizeof(sock_ts));
+ ts->tv_sec = sock_ts.tv_sec;
+ ts->tv_nsec = sock_ts.tv_nsec;
+}
+#endif /* defined(SO_TIMESTAMPNS_NEW) */
+#endif /* defined(SO_TIMESTAMPNS_OLD) */
+
+void get_timespec_from_so_timestamping(const struct cmsghdr *cmsg,
+ struct timespec *ts)
+{
+ int i;
+ struct scm_timestamping tss;
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SCM_TIMESTAMPING);
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(tss)));
+
+ memcpy(&tss, CMSG_DATA(cmsg), sizeof(tss));
+
+ for (i = 0; i < 3; ++i) {
+ if (tss.ts[i].tv_sec || tss.ts[i].tv_nsec) {
+ *ts = tss.ts[i];
+ return;
+ }
+ }
+ assert(!"All three entries in scm_timestamping are empty");
+}
+
+#ifdef SO_TIMESTAMPING_OLD
+void get_timespec_from_so_timestamping_old(const struct cmsghdr *cmsg,
+ struct timespec *ts)
+{
+ int i;
+ struct scm_old_timestamping tss;
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SO_TIMESTAMPING_OLD);
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(tss)));
+
+ memcpy(&tss, CMSG_DATA(cmsg), sizeof(tss));
+
+ for (i = 0; i < 3; ++i) {
+ if (tss.ts[i].tv_sec || tss.ts[i].tv_nsec) {
+ ts->tv_sec = tss.ts[i].tv_sec;
+ ts->tv_nsec = tss.ts[i].tv_nsec;
+ return;
+ }
+ }
+ assert(!"All three entries in scm_old_timestamping are empty");
+}
+
+#ifdef SO_TIMESTAMPING_NEW
+void get_timespec_from_so_timestamping_new(const struct cmsghdr *cmsg,
+ struct timespec *ts)
+{
+ int i;
+ struct scm_timestamping64 tss;
+ assert(cmsg->cmsg_level == SOL_SOCKET);
+ assert(cmsg->cmsg_type == SO_TIMESTAMPING_NEW);
+ assert(cmsg->cmsg_len == CMSG_LEN(sizeof(tss)));
+
+ memcpy(&tss, CMSG_DATA(cmsg), sizeof(tss));
+ for (i = 0; i < 3; ++i) {
+ if (tss.ts[i].tv_sec || tss.ts[i].tv_nsec) {
+ ts->tv_sec = tss.ts[i].tv_sec;
+ ts->tv_nsec = tss.ts[i].tv_nsec;
+ return;
+ }
+ }
+ assert(!"All three entries in scm_timestamp64 are empty");
+}
+#endif /* defined(SO_TIMESTAMPING_NEW) */
+#endif /* defined(SO_TIMESTAMPING_OLD) */
+
+void set_socket_option(int sock, int sockopt, int set_to)
{
socklen_t len;
- int val = on;
+ int val = set_to;
if (setsockopt(sock, SOL_SOCKET, sockopt, &val, sizeof(val)) < 0) {
int err = errno;
- fprintf(stderr, "Failed to setsockopt %d (%s): %s\n",
- sockopt, on ? "on" : "off", strerror(err));
+ fprintf(stderr, "Failed at setsockopt(%d, SOL_SOCKET, %d, %d): %s\n",
+ sock, sockopt, set_to, strerror(err));
exit(err);
}
+#ifdef SO_TIMESTAMPING_NEW
+ if (sockopt == SO_TIMESTAMPING_NEW) {
+ /* `getsockopt(sock, SOL_SOCKET, SO_TIMESTAMPING_NEW)` not implemented
+ * as of linux kernel v5.8-rc4.
+ */
+ return;
+ }
+#endif
+
len = sizeof(val);
val = -1;
if (getsockopt(sock, SOL_SOCKET, sockopt, &val, &len) < 0) {
int err = errno;
- fprintf(stderr, "Failed to getsockopt (%d): %s\n", sock, strerror(err));
+ fprintf(stderr, "Failed at getsockopt(%d, SOL_SOCKET, %d): %s\n",
+ sock, sockopt, strerror(err));
exit(err);
}
assert(len == sizeof(val));
- assert(val == on);
+ assert(val == set_to);
+}
+
+void child_steps(int sock, struct sockaddr_in addr, int run_old)
+{
+ /* Test 1: SO_TIMESTAMP */
+ send_current_time(sock, addr);
+
+ /* Test 2: SO_TIMESTAMPNS */
+ printf("Test 2: SO_TIMESTAMPNS\n");
+ set_socket_option(sock, SO_TIMESTAMPNS, 1);
+ receive_packet(sock, &get_timespec_from_so_timestampns);
+ set_socket_option(sock, SO_TIMESTAMPNS, 0);
+
+ /* Test 3: SO_TIMESTAMPING */
+ send_current_time(sock, addr);
+
+ if (!run_old) {
+ return;
+ }
+
+#ifdef SO_TIMESTAMP_OLD
+ if (SO_TIMESTAMP_OLD != SO_TIMESTAMP) {
+ /* Test 4a: SO_TIMESTAMP_OLD */
+ printf("Test 4a: SO_TIMESTAMP_OLD\n");
+ set_socket_option(sock, SO_TIMESTAMP_OLD, 1);
+ receive_packet(sock, &get_timespec_from_so_timestamp_old);
+ set_socket_option(sock, SO_TIMESTAMP_OLD, 0);
+ }
+#ifdef SO_TIMESTAMP_NEW
+ else {
+ /* Test 4b: SO_TIMESTAMP_NEW */
+ printf("Test 4b: SO_TIMESTAMP_NEW\n");
+ set_socket_option(sock, SO_TIMESTAMP_NEW, 1);
+ receive_packet(sock, &get_timespec_from_so_timestamp_new);
+ set_socket_option(sock, SO_TIMESTAMP_NEW, 0);
+ }
+#endif /* defined(SO_TIMESTAMP_NEW) */
+#endif /* defined(SO_TIMESTAMP_OLD) */
+
+#ifdef SO_TIMESTAMPNS_OLD
+ if (SO_TIMESTAMPNS_OLD != SO_TIMESTAMPNS) {
+ /* Test 5a: SO_TIMESTAMPNS_OLD */
+ send_current_time(sock, addr);
+ }
+#ifdef SO_TIMESTAMPNS_NEW
+ else {
+ /* Test 5b: SO_TIMESTAMPNS_NEW */
+ send_current_time(sock, addr);
+ }
+#endif /* defined(SO_TIMESTAMPNS_NEW) */
+#endif /* defined(SO_TIMESTAMPNS_OLD) */
+
+#ifdef SO_TIMESTAMPING_OLD
+ if (SO_TIMESTAMPING_OLD != SO_TIMESTAMPING) {
+ /* Test 6a: SO_TIMESTAMPING_OLD */
+ printf("Test 6a: SO_TIMESTAMPING_OLD\n");
+ set_socket_option(sock, SO_TIMESTAMPING_OLD, so_timestamping_flags);
+ receive_packet(sock, &get_timespec_from_so_timestamping_old);
+ set_socket_option(sock, SO_TIMESTAMPING_OLD, 0);
+ }
+#ifdef SO_TIMESTAMPING_NEW
+ else {
+ /* Test 6b: SO_TIMESTAMPING_NEW */
+ printf("Test 6b: SO_TIMESTAMPING_NEW\n");
+ set_socket_option(sock, SO_TIMESTAMPING_NEW, so_timestamping_flags);
+ receive_packet(sock, &get_timespec_from_so_timestamping_new);
+ set_socket_option(sock, SO_TIMESTAMPING_NEW, 0);
+ }
+#endif /* defined(SO_TIMESTAMPING_NEW) */
+#endif /* defined(SO_TIMESTAMPING_OLD) */
+}
+
+void parent_steps(int sock, struct sockaddr_in addr, int run_old)
+{
+ /* Test 1: SO_TIMESTAMP */
+ printf("Test 1: SO_TIMESTAMP\n");
+ set_socket_option(sock, SO_TIMESTAMP, 1);
+ receive_packet(sock, &get_timespec_from_so_timestamp);
+ set_socket_option(sock, SO_TIMESTAMP, 0);
+
+ /* Test 2: SO_TIMESTAMPNS */
+ send_current_time(sock, addr);
+
+ /* Test 3: SO_TIMESTAMPING */
+ printf("Test 3: SO_TIMESTAMPING\n");
+ set_socket_option(sock, SO_TIMESTAMPING, so_timestamping_flags);
+ receive_packet(sock, &get_timespec_from_so_timestamping);
+ set_socket_option(sock, SO_TIMESTAMPING, 0);
+
+ if (!run_old) {
+ return;
+ }
+
+#ifdef SO_TIMESTAMP_OLD
+ if (SO_TIMESTAMP_OLD != SO_TIMESTAMP) {
+ /* Test 4a: SO_TIMESTAMP_OLD */
+ send_current_time(sock, addr);
+ }
+#ifdef SO_TIMESTAMP_NEW
+ else {
+ /* Test 4b: SO_TIMESTAMP_NEW */
+ send_current_time(sock, addr);
+ }
+#endif /* defined(SO_TIMESTAMP_NEW) */
+#endif /* defined(SO_TIMESTAMP_OLD) */
+
+#ifdef SO_TIMESTAMPNS_OLD
+ if (SO_TIMESTAMPNS_OLD != SO_TIMESTAMPNS) {
+ /* Test 5a: SO_TIMESTAMPNS_OLD */
+ printf("Test 5a: SO_TIMESTAMPNS_OLD\n");
+ set_socket_option(sock, SO_TIMESTAMPNS_OLD, 1);
+ receive_packet(sock, &get_timespec_from_so_timestampns_old);
+ set_socket_option(sock, SO_TIMESTAMPNS_OLD, 0);
+ }
+#ifdef SO_TIMESTAMPNS_NEW
+ else {
+ /* Test 5b: SO_TIMESTAMPNS_NEW */
+ printf("Test 5b: SO_TIMESTAMPNS_NEW\n");
+ set_socket_option(sock, SO_TIMESTAMPNS_NEW, 1);
+ receive_packet(sock, &get_timespec_from_so_timestampns_new);
+ set_socket_option(sock, SO_TIMESTAMPNS_NEW, 0);
+ }
+#endif /* defined(SO_TIMESTAMPNS_NEW) */
+#endif /* defined(SO_TIMESTAMPNS_OLD) */
+
+#ifdef SO_TIMESTAMPING_OLD
+ if (SO_TIMESTAMPING_OLD != SO_TIMESTAMPING) {
+ /* Test 6a: SO_TIMESTAMPING_OLD */
+ send_current_time(sock, addr);
+ }
+#ifdef SO_TIMESTAMPING_NEW
+ else {
+ /* Test 6b: SO_TIMESTAMPING_NEW */
+ send_current_time(sock, addr);
+ }
+#endif /* defined(SO_TIMESTAMPING_NEW) */
+#endif /* defined(SO_TIMESTAMPING_OLD) */
}
int main(int argc, char **argv)
{
int parent_sock, child_sock;
struct sockaddr_in parent_sockaddr, child_sockaddr;
- int pid;
+ int pid, run_old;
struct timeval tv = {0, 0};
gettimeofday(&tv, NULL);
+ /* Too close to y2038 old systems may not work. */
+ run_old = tv.tv_sec < 0x7fffff00;
+
parent_sock = create_udp_socket(&parent_sockaddr);
child_sock = create_udp_socket(&child_sockaddr);
printf("Parent sock bound to port %d\nChild sock bound to port %d\n",
parent_sockaddr.sin_port, child_sockaddr.sin_port);
- if ((pid = fork()) == -2) {
+ if ((pid = fork()) < 0) {
fprintf(stderr, "SKIPPED. Failed to fork: %s\n", strerror(errno));
} else if (pid == 0) {
- close(child_sock);
-
- /* Test 1: SO_TIMESTAMP */
- send_current_time(parent_sock, child_sockaddr);
-
- if (tv.tv_sec > 0x7fffff00) {
- /* Too close to y2038 problem, old system may not work. */
- close(parent_sock);
- return 0;
- }
-
-#ifdef SO_TIMESTAMP_OLD
- if (SO_TIMESTAMP_OLD != SO_TIMESTAMP) {
- /* Test 2a: SO_TIMESTAMP_OLD */
- set_socket_option(parent_sock, SO_TIMESTAMP_OLD, 1);
- receive_packet(parent_sock, &get_timeval_from_so_timestamp_old);
- set_socket_option(parent_sock, SO_TIMESTAMP_OLD, 0);
- }
-#ifdef SO_TIMESTAMP_NEW
- else {
- /* Test 2b: SO_TIMESTAMP_NEW */
- set_socket_option(parent_sock, SO_TIMESTAMP_NEW, 1);
- receive_packet(parent_sock, &get_timeval_from_so_timestamp_new);
- set_socket_option(parent_sock, SO_TIMESTAMP_NEW, 0);
- }
-#endif /* defined(SO_TIMESTAMP_NEW) */
-#endif /* defined(SO_TIMESTAMP_OLD) */
-
close(parent_sock);
+ child_steps(child_sock, parent_sockaddr, run_old);
+ close(child_sock);
} else {
int child_status;
- close(parent_sock);
-
- /* Test 1: SO_TIMESTAMP */
- set_socket_option(child_sock, SO_TIMESTAMP, 1);
- receive_packet(child_sock, &get_timeval_from_so_timestamp);
- set_socket_option(child_sock, SO_TIMESTAMP, 0);
-
- if (tv.tv_sec > 0x7fffff00) {
- /* Too close to y2038 problem, old system may not work. */
- close(child_sock);
- return 0;
- }
-
-#ifdef SO_TIMESTAMP_OLD
- if (SO_TIMESTAMP_OLD != SO_TIMESTAMP) {
- /* Test 2a: SO_TIMESTAMP_OLD */
- send_current_time(child_sock, parent_sockaddr);
- }
-#ifdef SO_TIMESTAMP_NEW
- else {
- /* Test 2b: SO_TIMESTAMP_NEW */
- send_current_time(child_sock, parent_sockaddr);
- }
-#endif /* defined(SO_TIMESTAMP_NEW) */
-#endif /* defined(SO_TIMESTAMP_OLD) */
close(child_sock);
+ parent_steps(parent_sock, child_sockaddr, run_old);
+ close(parent_sock);
if (waitpid(pid, &child_status, 0) < 0) {
int err = errno;
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 5/6] thunk: supports flexible arrays
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
` (3 preceding siblings ...)
2020-07-23 0:19 ` [PATCH 4/6] linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING Shu-Chun Weng
@ 2020-07-23 0:19 ` Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 6/6] linux-user: Add support for SIOCETHTOOL ioctl Shu-Chun Weng
2020-08-05 23:22 ` [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
6 siblings, 0 replies; 11+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:19 UTC (permalink / raw)
To: qemu-devel; +Cc: Shu-Chun Weng, laurent, riku.voipio
Flexible arrays may appear in the last field of a struct and are heavily
used in the ioctl(SIOCETHTOOL) system call on Linux. E.g.
struct ethtool_regs {
__u32 cmd;
__u32 version; /* driver-specific, indicates different chips/revs */
__u32 len; /* bytes */
__u8 data[0];
};
where number of elements in `data` is specified in `len`. It is translated
into:
STRUCT(ethtool_regs,
TYPE_INT, /* cmd */
TYPE_INT, /* version */
TYPE_INT, /* len */
MK_FLEXIBLE_ARRAY(TYPE_CHAR, 2)) /* data[0]: len */
where the "2" passed to `MK_FLEXIBLE_ARRAY` means the number of element
is specified by field number 2 (0-index).
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
include/exec/user/thunk.h | 20 +++++
thunk.c | 151 +++++++++++++++++++++++++++++++++++++-
2 files changed, 169 insertions(+), 2 deletions(-)
diff --git a/include/exec/user/thunk.h b/include/exec/user/thunk.h
index 7992475c9f..080d84e806 100644
--- a/include/exec/user/thunk.h
+++ b/include/exec/user/thunk.h
@@ -39,12 +39,19 @@ typedef enum argtype {
TYPE_ARRAY,
TYPE_STRUCT,
TYPE_OLDDEVT,
+ TYPE_FLEXIBLE_ARRAY,
} argtype;
#define MK_PTR(type) TYPE_PTR, type
#define MK_ARRAY(type, size) TYPE_ARRAY, size, type
#define MK_STRUCT(id) TYPE_STRUCT, id
+/* Should only appear as the last element of a TYPE_STRUCT. `len_field_idx` is
+ * the index into the fields in the enclosing struct that specify the length of
+ * the flexibly array. The length field MUST be a TYPE_INT field. */
+#define MK_FLEXIBLE_ARRAY(type, len_field_idx) \
+ TYPE_FLEXIBLE_ARRAY, len_field_idx, type
+
#define THUNK_TARGET 0
#define THUNK_HOST 1
@@ -55,6 +62,8 @@ typedef struct {
int *field_offsets[2];
/* special handling */
void (*convert[2])(void *dst, const void *src);
+ int (*thunk_size[2])(const void *src);
+
int size[2];
int align[2];
const char *name;
@@ -75,6 +84,11 @@ const argtype *thunk_convert(void *dst, const void *src,
const argtype *type_ptr, int to_host);
const argtype *thunk_print(void *arg, const argtype *type_ptr);
+bool thunk_type_has_flexible_array(const argtype *type_ptr);
+/* thunk_type_size but can handle TYPE_FLEXIBLE_ARRAY */
+int thunk_type_size_with_src(const void *src, const argtype *type_ptr,
+ int is_host);
+
extern StructEntry *struct_entries;
int thunk_type_size_array(const argtype *type_ptr, int is_host);
@@ -137,6 +151,10 @@ static inline int thunk_type_size(const argtype *type_ptr, int is_host)
case TYPE_STRUCT:
se = struct_entries + type_ptr[1];
return se->size[is_host];
+ case TYPE_FLEXIBLE_ARRAY:
+ /* Flexible arrays do not count toward sizeof(). Users of structures
+ * containing them need to calculate it themselves. */
+ return 0;
default:
g_assert_not_reached();
}
@@ -187,6 +205,8 @@ static inline int thunk_type_align(const argtype *type_ptr, int is_host)
case TYPE_STRUCT:
se = struct_entries + type_ptr[1];
return se->align[is_host];
+ case TYPE_FLEXIBLE_ARRAY:
+ return thunk_type_align_array(type_ptr + 2, is_host);
default:
g_assert_not_reached();
}
diff --git a/thunk.c b/thunk.c
index c5d9719747..7b89332712 100644
--- a/thunk.c
+++ b/thunk.c
@@ -50,6 +50,8 @@ static inline const argtype *thunk_type_next(const argtype *type_ptr)
return thunk_type_next_ptr(type_ptr + 1);
case TYPE_STRUCT:
return type_ptr + 1;
+ case TYPE_FLEXIBLE_ARRAY:
+ return thunk_type_next_ptr(type_ptr + 1);
default:
return NULL;
}
@@ -122,6 +124,34 @@ void thunk_register_struct_direct(int id, const char *name,
se->name = name;
}
+static const argtype *
+thunk_convert_flexible_array(void *dst, const void *src,
+ const uint8_t *dst_struct,
+ const uint8_t *src_struct, const argtype *type_ptr,
+ const StructEntry *se, int to_host) {
+ int len_field_idx, dst_size, src_size, i;
+ uint32_t array_length;
+ uint8_t *d;
+ const uint8_t *s;
+
+ assert(*type_ptr == TYPE_FLEXIBLE_ARRAY);
+ type_ptr++;
+ len_field_idx = *type_ptr++;
+ array_length =
+ *(const uint32_t *)(to_host ?
+ dst_struct + se->field_offsets[1][len_field_idx] :
+ src_struct + se->field_offsets[0][len_field_idx]);
+ dst_size = thunk_type_size(type_ptr, to_host);
+ src_size = thunk_type_size(type_ptr, to_host);
+ d = dst;
+ s = src;
+ for (i = 0; i < array_length; i++) {
+ thunk_convert(d, s, type_ptr, to_host);
+ d += dst_size;
+ s += src_size;
+ }
+ return thunk_type_next(type_ptr);
+}
/* now we can define the main conversion functions */
const argtype *thunk_convert(void *dst, const void *src,
@@ -246,7 +276,7 @@ const argtype *thunk_convert(void *dst, const void *src,
assert(*type_ptr < max_struct_entries);
se = struct_entries + *type_ptr++;
- if (se->convert[0] != NULL) {
+ if (se->convert[to_host] != NULL) {
/* specific conversion is needed */
(*se->convert[to_host])(dst, src);
} else {
@@ -256,7 +286,18 @@ const argtype *thunk_convert(void *dst, const void *src,
src_offsets = se->field_offsets[1 - to_host];
d = dst;
s = src;
- for(i = 0;i < se->nb_fields; i++) {
+ for(i = 0; i < se->nb_fields; i++) {
+ if (*field_types == TYPE_FLEXIBLE_ARRAY) {
+ field_types = thunk_convert_flexible_array(
+ d + dst_offsets[i],
+ s + src_offsets[i],
+ d,
+ s,
+ field_types,
+ se,
+ to_host);
+ continue;
+ }
field_types = thunk_convert(d + dst_offsets[i],
s + src_offsets[i],
field_types, to_host);
@@ -264,6 +305,11 @@ const argtype *thunk_convert(void *dst, const void *src,
}
}
break;
+ case TYPE_FLEXIBLE_ARRAY:
+ fprintf(stderr,
+ "Invalid flexible array (type 0x%x) outside of a structure\n",
+ type);
+ break;
default:
fprintf(stderr, "Invalid type 0x%x\n", type);
break;
@@ -271,6 +317,45 @@ const argtype *thunk_convert(void *dst, const void *src,
return type_ptr;
}
+static const argtype *
+thunk_print_flexible_array(void *arg, const uint8_t *arg_struct,
+ const argtype *type_ptr, const StructEntry *se) {
+ int array_length, len_field_idx, arg_size, i;
+ uint8_t *a;
+ int is_string = 0;
+
+ assert(*type_ptr == TYPE_FLEXIBLE_ARRAY);
+ type_ptr++;
+ len_field_idx = *type_ptr++;
+
+ array_length = tswap32(
+ *(const uint32_t *)(arg_struct + se->field_offsets[0][len_field_idx]));
+ arg_size = thunk_type_size(type_ptr, 0);
+ a = arg;
+
+ if (*type_ptr == TYPE_CHAR) {
+ qemu_log("\"");
+ is_string = 1;
+ } else {
+ qemu_log("[");
+ }
+
+ for (i = 0; i < array_length; i++) {
+ if (i > 0 && !is_string) {
+ qemu_log(",");
+ }
+ thunk_print(a, type_ptr);
+ a += arg_size;
+ }
+
+ if (is_string) {
+ qemu_log("\"");
+ } else {
+ qemu_log("]");
+ }
+ return thunk_type_next(type_ptr);
+}
+
const argtype *thunk_print(void *arg, const argtype *type_ptr)
{
int type;
@@ -414,17 +499,79 @@ const argtype *thunk_print(void *arg, const argtype *type_ptr)
if (i > 0) {
qemu_log(",");
}
+ if (*field_types == TYPE_FLEXIBLE_ARRAY) {
+ field_types = thunk_print_flexible_array(
+ a + arg_offsets[i], a, field_types, se);
+ continue;
+ }
field_types = thunk_print(a + arg_offsets[i], field_types);
}
qemu_log("}");
}
break;
+ case TYPE_FLEXIBLE_ARRAY:
+ fprintf(stderr,
+ "Invalid flexible array (type 0x%x) outside of a structure\n",
+ type);
+ break;
default:
g_assert_not_reached();
}
return type_ptr;
}
+bool thunk_type_has_flexible_array(const argtype *type_ptr) {
+ int i;
+ const StructEntry *se;
+ const argtype *field_types;
+ if (*type_ptr != TYPE_STRUCT) {
+ return false;
+ }
+ se = struct_entries + type_ptr[1];
+ field_types = se->field_types;
+ for (i = 0; i < se->nb_fields; i++) {
+ if (*field_types == TYPE_FLEXIBLE_ARRAY) {
+ return true;
+ }
+ field_types = thunk_type_next(type_ptr);
+ }
+ return false;
+}
+
+int thunk_type_size_with_src(const void *src, const argtype *type_ptr,
+ int is_host)
+{
+ switch(*type_ptr) {
+ case TYPE_STRUCT: {
+ int i;
+ const StructEntry *se = struct_entries + type_ptr[1];
+ const argtype *field_types;
+ if (se->thunk_size[is_host] != NULL) {
+ return (*se->thunk_size[is_host])(src);
+ }
+
+ field_types = se->field_types;
+ for (i = 0; i < se->nb_fields; i++) {
+ if (*field_types == TYPE_FLEXIBLE_ARRAY) {
+ uint32_t array_length = *(const uint32_t *)(
+ (const uint8_t*) src +
+ se->field_offsets[is_host][field_types[1]]);
+ if (!is_host) {
+ array_length = tswap32(array_length);
+ }
+ return se->size[is_host] +
+ array_length *
+ thunk_type_size(field_types + 2, is_host);
+ }
+ field_types = thunk_type_next(type_ptr);
+ }
+ return se->size[is_host];
+ }
+ default:
+ return thunk_type_size(type_ptr, is_host);
+ }
+}
+
/* from em86 */
/* Utility function: Table-driven functions to translate bitmasks
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 6/6] linux-user: Add support for SIOCETHTOOL ioctl
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
` (4 preceding siblings ...)
2020-07-23 0:19 ` [PATCH 5/6] thunk: supports flexible arrays Shu-Chun Weng
@ 2020-07-23 0:19 ` Shu-Chun Weng
2020-08-05 23:22 ` [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
6 siblings, 0 replies; 11+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:19 UTC (permalink / raw)
To: qemu-devel; +Cc: Shu-Chun Weng, laurent, riku.voipio, alex.bennee
The ioctl numeric values are platform-independent and determined by
the file include/uapi/linux/sockios.h in Linux kernel source code:
#define SIOCETHTOOL 0x8946
These ioctls get (or set) various structures pointed by the field
ifr_data in the structure ifreq depending on the first 4 bytes of the
memory region.
This change clones the ioctl framework into ethtool-specific dispatch
logic in its own file. A number of definitions previously only visible
in syscall.c are thus exported to syscall_defs.h to be used in the new
files.
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
linux-user/Makefile.objs | 3 +-
linux-user/ethtool.c | 819 ++++++++++++++++++++++++++++++++++
linux-user/ethtool.h | 19 +
linux-user/ethtool_entries.h | 107 +++++
linux-user/ioctls.h | 2 +
linux-user/qemu.h | 1 +
linux-user/syscall.c | 35 +-
linux-user/syscall_defs.h | 12 +
linux-user/syscall_types.h | 277 ++++++++++++
tests/tcg/multiarch/ethtool.c | 417 +++++++++++++++++
10 files changed, 1680 insertions(+), 12 deletions(-)
create mode 100644 linux-user/ethtool.c
create mode 100644 linux-user/ethtool.h
create mode 100644 linux-user/ethtool_entries.h
create mode 100644 tests/tcg/multiarch/ethtool.c
diff --git a/linux-user/Makefile.objs b/linux-user/Makefile.objs
index 1940910a73..971d43173a 100644
--- a/linux-user/Makefile.objs
+++ b/linux-user/Makefile.objs
@@ -1,7 +1,8 @@
obj-y = main.o syscall.o strace.o mmap.o signal.o \
elfload.o linuxload.o uaccess.o uname.o \
safe-syscall.o $(TARGET_ABI_DIR)/signal.o \
- $(TARGET_ABI_DIR)/cpu_loop.o exit.o fd-trans.o
+ $(TARGET_ABI_DIR)/cpu_loop.o exit.o fd-trans.o \
+ ethtool.o
obj-$(TARGET_HAS_BFLT) += flatload.o
obj-$(TARGET_I386) += vm86.o
diff --git a/linux-user/ethtool.c b/linux-user/ethtool.c
new file mode 100644
index 0000000000..cb134e7c9b
--- /dev/null
+++ b/linux-user/ethtool.c
@@ -0,0 +1,819 @@
+/*
+ * Linux ioctl system call SIOCETHTOOL requests
+ *
+ * Copyright (c) 2020 Shu-Chun Weng
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+#include "qemu/osdep.h"
+#include <stdio.h>
+#include <linux/ethtool.h>
+#include <linux/if.h>
+#include <linux/sockios.h>
+#include <linux/unistd.h>
+#include "ethtool.h"
+#include "qemu.h"
+#include "syscall_defs.h"
+
+/* Non-standard ethtool structure definitions. */
+/* struct ethtool_rxnfc {
+ * __u32 cmd;
+ * __u32 flow_type;
+ * __u64 data;
+ * struct ethtool_rx_flow_spec fs;
+ * union {
+ * __u32 rule_cnt;
+ * __u32 rss_context;
+ * };
+ * __u32 rule_locs[0];
+ * };
+ *
+ * Originally defined for ETHTOOL_{G,S}RXFH with only the cmd, flow_type and
+ * data members. For other commands, dedicated standard structure definitions
+ * are listed in syscall_types.h.
+ */
+static void host_to_target_ethtool_rxnfc_get_set_rxfh(void *dst,
+ const void *src)
+{
+ static const argtype ethtool_rx_flow_spec_argtype[] = {
+ MK_STRUCT(STRUCT_ethtool_rx_flow_spec), TYPE_NULL };
+ struct ethtool_rxnfc *target = dst;
+ const struct ethtool_rxnfc *host = src;
+
+ target->cmd = tswap32(host->cmd);
+ target->flow_type = tswap32(host->flow_type);
+ target->data = tswap64(host->data);
+
+ if (host->cmd == ETHTOOL_SRXFH) {
+ /* struct ethtool_rxnfc was originally defined for ETHTOOL_{G,S}RXFH
+ * with only the cmd, flow_type and data members. Guest program might
+ * still be using that definition.
+ */
+ return;
+ }
+ if (host->cmd != ETHTOOL_GRXFH) {
+ fprintf(stderr, "host_to_target_ethtool_rxnfc_get_set_rxfh called with "
+ "command 0x%x which is not ETHTOOL_SRXFH or ETHTOOL_GRXFH\n",
+ host->cmd);
+ }
+ if ((host->flow_type & FLOW_RSS) == 0) {
+ return;
+ }
+ /* If `FLOW_RSS` was requested then guest program must be using the new
+ * definition.
+ */
+ thunk_convert(&target->fs, &host->fs, ethtool_rx_flow_spec_argtype,
+ THUNK_TARGET);
+ target->rule_cnt = tswap32(host->rule_cnt);
+}
+
+static void target_to_host_ethtool_rxnfc_get_set_rxfh(void *dst,
+ const void *src)
+{
+ static const argtype ethtool_rx_flow_spec_argtype[] = {
+ MK_STRUCT(STRUCT_ethtool_rx_flow_spec), TYPE_NULL };
+ struct ethtool_rxnfc *host = dst;
+ const struct ethtool_rxnfc *target = src;
+
+ host->cmd = tswap32(target->cmd);
+ host->flow_type = tswap32(target->flow_type);
+ host->data = tswap64(target->data);
+
+ if (host->cmd == ETHTOOL_SRXFH) {
+ /* struct ethtool_rxnfc was originally defined for ETHTOOL_{G,S}RXFH
+ * with only the cmd, flow_type and data members. Guest program might
+ * still be using that definition.
+ */
+ return;
+ }
+ if (host->cmd != ETHTOOL_GRXFH) {
+ fprintf(stderr, "target_to_host_ethtool_rxnfc_get_set_rxfh called with "
+ "command 0x%x which is not ETHTOOL_SRXFH or ETHTOOL_GRXFH\n",
+ host->cmd);
+ }
+ if ((host->flow_type & FLOW_RSS) == 0) {
+ return;
+ }
+ /* If `FLOW_RSS` was requested then guest program must be using the new
+ * definition.
+ */
+ thunk_convert(&host->fs, &target->fs, ethtool_rx_flow_spec_argtype,
+ THUNK_HOST);
+ host->rule_cnt = tswap32(target->rule_cnt);
+}
+
+static int target_ethtool_rxnfc_get_set_rxfh_size(const void *src)
+{
+ const struct ethtool_rxnfc *target = src;
+ int cmd = tswap32(target->cmd);
+ if (cmd == ETHTOOL_SRXFH ||
+ (cmd == ETHTOOL_GRXFH &&
+ (tswap32(target->flow_type) & FLOW_RSS) == 0)) {
+ return 16;
+ }
+ return sizeof(struct ethtool_rxnfc);
+}
+
+static int host_ethtool_rxnfc_get_set_rxfh_size(const void *src)
+{
+ const struct ethtool_rxnfc *host = src;
+ if (host->cmd == ETHTOOL_SRXFH ||
+ (host->cmd == ETHTOOL_GRXFH && (host->flow_type & FLOW_RSS) == 0)) {
+ return 16;
+ }
+ return sizeof(struct ethtool_rxnfc);
+}
+
+const StructEntry struct_ethtool_rxnfc_get_set_rxfh_def = {
+ .convert = {
+ host_to_target_ethtool_rxnfc_get_set_rxfh,
+ target_to_host_ethtool_rxnfc_get_set_rxfh },
+ .thunk_size = {
+ target_ethtool_rxnfc_get_set_rxfh_size,
+ host_ethtool_rxnfc_get_set_rxfh_size },
+ .size = { 16, 16 },
+ .align = {
+ __alignof__(struct ethtool_rxnfc),
+ __alignof__(struct ethtool_rxnfc) },
+};
+
+/* struct ethtool_sset_info {
+ * __u32 cmd;
+ * __u32 reserved;
+ * __u64 sset_mask;
+ * __u32 data[0];
+ * };
+ *
+ * `sset_mask` is a bitmask of string sets. `data` is the buffer for string set
+ * sizes, containing number of 1s in `sset_mask`'s binary representation number
+ * of 4-byte entries.
+ *
+ * Since all fields are fixed-width and number of 1s in `sset_mask` does not
+ * change between architectures, host-to-target and target-to-host are
+ * identical.
+ */
+static void convert_ethtool_sset_info(void *dst, const void *src)
+{
+ int i, set_count;
+ struct ethtool_sset_info *dst_sset_info = dst;
+ const struct ethtool_sset_info *src_sset_info = src;
+
+ dst_sset_info->cmd = tswap32(src_sset_info->cmd);
+ dst_sset_info->sset_mask = tswap64(src_sset_info->sset_mask);
+
+ set_count = ctpop64(src_sset_info->sset_mask);
+ for (i = 0; i < set_count; ++i) {
+ dst_sset_info->data[i] = tswap32(src_sset_info->data[i]);
+ }
+}
+
+static int ethtool_sset_info_size(const void *src)
+{
+ const struct ethtool_sset_info *src_sset_info = src;
+ return sizeof(struct ethtool_sset_info) +
+ ctpop64(src_sset_info->sset_mask) * sizeof(src_sset_info->data[0]);
+}
+
+const StructEntry struct_ethtool_sset_info_def = {
+ .convert = {
+ convert_ethtool_sset_info, convert_ethtool_sset_info },
+ .thunk_size = {
+ ethtool_sset_info_size, ethtool_sset_info_size },
+ .size = {
+ sizeof(struct ethtool_sset_info),
+ sizeof(struct ethtool_sset_info) },
+ .align = {
+ __alignof__(struct ethtool_sset_info),
+ __alignof__(struct ethtool_sset_info) },
+};
+
+/* struct ethtool_rxfh {
+ * __u32 cmd;
+ * __u32 rss_context;
+ * __u32 indir_size;
+ * __u32 key_size;
+ * __u8 hfunc;
+ * __u8 rsvd8[3];
+ * __u32 rsvd32;
+ * __u32 rss_config[0];
+ * };
+ *
+ * `rss_config`: indirection table of `indir_size` __u32 elements, followed by
+ * hash key of `key_size` bytes.
+ *
+ * `indir_size` could be ETH_RXFH_INDIR_NO_CHANGE when `cmd` is ETHTOOL_SRSSH
+ * and there would be no indircetion table in `rss_config`.
+ */
+static void convert_ethtool_rxfh_header(void *dst, const void *src)
+{
+ struct ethtool_rxfh *dst_rxfh = dst;
+ const struct ethtool_rxfh *src_rxfh = src;
+
+ dst_rxfh->cmd = tswap32(src_rxfh->cmd);
+ dst_rxfh->rss_context = tswap32(src_rxfh->rss_context);
+ dst_rxfh->indir_size = tswap32(src_rxfh->indir_size);
+ dst_rxfh->key_size = tswap32(src_rxfh->key_size);
+ dst_rxfh->hfunc = src_rxfh->hfunc;
+ dst_rxfh->rsvd8[0] = src_rxfh->rsvd8[0];
+ dst_rxfh->rsvd8[1] = src_rxfh->rsvd8[1];
+ dst_rxfh->rsvd8[2] = src_rxfh->rsvd8[2];
+ dst_rxfh->rsvd32 = tswap32(src_rxfh->rsvd32);
+}
+
+static void convert_ethtool_rxfh_rss_config(
+ void *dst, const void *src, uint32_t indir_size, uint32_t key_size) {
+ uint32_t *dst_rss_config = (uint32_t *)dst;
+ const uint32_t *src_rss_config = (const uint32_t *)src;
+ int i;
+ for (i = 0; i < indir_size; ++i) {
+ dst_rss_config[i] = tswap32(src_rss_config[i]);
+ }
+ if (key_size > 0) {
+ memcpy(dst_rss_config + indir_size,
+ src_rss_config + indir_size,
+ key_size);
+ }
+}
+
+static void host_to_target_ethtool_rxfh(void *dst, const void *src)
+{
+ struct ethtool_rxfh *target = dst;
+ const struct ethtool_rxfh *host = src;
+
+ convert_ethtool_rxfh_header(dst, src);
+
+ const uint32_t indir_size =
+ host->cmd == ETHTOOL_SRSSH &&
+ host->indir_size == ETH_RXFH_INDIR_NO_CHANGE ?
+ 0 :
+ host->indir_size;
+ convert_ethtool_rxfh_rss_config(target->rss_config, host->rss_config,
+ indir_size, host->key_size);
+}
+
+static void target_to_host_ethtool_rxfh(void *dst, const void *src)
+{
+ struct ethtool_rxfh *host = dst;
+ const struct ethtool_rxfh *target = src;
+
+ convert_ethtool_rxfh_header(dst, src);
+
+ const uint32_t indir_size =
+ host->cmd == ETHTOOL_SRSSH &&
+ host->indir_size == ETH_RXFH_INDIR_NO_CHANGE ?
+ 0 :
+ host->indir_size;
+ convert_ethtool_rxfh_rss_config(host->rss_config, target->rss_config,
+ indir_size, host->key_size);
+}
+
+static int target_ethtool_rxfh_size(const void *src)
+{
+ const struct ethtool_rxfh *target = src;
+ if (tswap32(target->cmd) == ETHTOOL_SRSSH &&
+ tswap32(target->indir_size) == ETH_RXFH_INDIR_NO_CHANGE) {
+ return sizeof(struct ethtool_rxfh) + tswap32(target->key_size);
+ }
+ return sizeof(struct ethtool_rxfh) +
+ tswap32(target->indir_size) * sizeof(target->rss_config[0]) +
+ tswap32(target->key_size);
+}
+
+static int host_ethtool_rxfh_size(const void *src)
+{
+ const struct ethtool_rxfh *host = src;
+ if (host->cmd == ETHTOOL_SRSSH &&
+ host->indir_size == ETH_RXFH_INDIR_NO_CHANGE) {
+ return sizeof(struct ethtool_rxfh) + host->key_size;
+ }
+ return sizeof(struct ethtool_rxfh) +
+ host->indir_size * sizeof(host->rss_config[0]) +
+ host->key_size;
+}
+
+const StructEntry struct_ethtool_rxfh_def = {
+ .convert = {
+ host_to_target_ethtool_rxfh, target_to_host_ethtool_rxfh },
+ .thunk_size = {
+ target_ethtool_rxfh_size, host_ethtool_rxfh_size },
+ .size = {
+ sizeof(struct ethtool_rxfh), sizeof(struct ethtool_rxfh) },
+ .align = {
+ __alignof__(struct ethtool_rxfh), __alignof__(struct ethtool_rxfh) },
+};
+
+/* struct ethtool_link_settings {
+ * __u32 cmd;
+ * __u32 speed;
+ * __u8 duplex;
+ * __u8 port;
+ * __u8 phy_address;
+ * __u8 autoneg;
+ * __u8 mdio_support;
+ * __u8 eth_tp_mdix;
+ * __u8 eth_tp_mdix_ctrl;
+ * __s8 link_mode_masks_nwords;
+ * __u8 transceiver;
+ * __u8 reserved1[3];
+ * __u32 reserved[7];
+ * __u32 link_mode_masks[0];
+ * };
+ *
+ * layout of link_mode_masks fields:
+ * __u32 map_supported[link_mode_masks_nwords];
+ * __u32 map_advertising[link_mode_masks_nwords];
+ * __u32 map_lp_advertising[link_mode_masks_nwords];
+ *
+ * `link_mode_masks_nwords` can be negative when returning from kernel if the
+ * provided request size is not supported.
+ */
+
+static void host_to_target_ethtool_link_settings(void *dst, const void *src)
+{
+ int i;
+ struct ethtool_link_settings *target = dst;
+ const struct ethtool_link_settings *host = src;
+
+ target->cmd = tswap32(host->cmd);
+ target->speed = tswap32(host->speed);
+ target->duplex = host->duplex;
+ target->port = host->port;
+ target->phy_address = host->phy_address;
+ target->autoneg = host->autoneg;
+ target->mdio_support = host->mdio_support;
+ target->eth_tp_mdix = host->eth_tp_mdix;
+ target->eth_tp_mdix_ctrl = host->eth_tp_mdix_ctrl;
+ target->link_mode_masks_nwords = host->link_mode_masks_nwords;
+ target->transceiver = host->transceiver;
+ for (i = 0; i < 3; ++i) {
+ target->reserved1[i] = host->reserved1[i];
+ }
+ for (i = 0; i < 7; ++i) {
+ target->reserved[i] = tswap32(host->reserved[i]);
+ }
+
+ if (host->link_mode_masks_nwords > 0) {
+ for (i = 0; i < host->link_mode_masks_nwords * 3; ++i) {
+ target->link_mode_masks[i] = tswap32(host->link_mode_masks[i]);
+ }
+ }
+}
+
+static void target_to_host_ethtool_link_settings(void *dst, const void *src)
+{
+ int i;
+ struct ethtool_link_settings *host = dst;
+ const struct ethtool_link_settings *target = src;
+
+ host->cmd = tswap32(target->cmd);
+ host->speed = tswap32(target->speed);
+ host->duplex = target->duplex;
+ host->port = target->port;
+ host->phy_address = target->phy_address;
+ host->autoneg = target->autoneg;
+ host->mdio_support = target->mdio_support;
+ host->eth_tp_mdix = target->eth_tp_mdix;
+ host->eth_tp_mdix_ctrl = target->eth_tp_mdix_ctrl;
+ host->link_mode_masks_nwords = target->link_mode_masks_nwords;
+ host->transceiver = target->transceiver;
+ for (i = 0; i < 3; ++i) {
+ host->reserved1[i] = target->reserved1[i];
+ }
+ for (i = 0; i < 7; ++i) {
+ host->reserved[i] = tswap32(target->reserved[i]);
+ }
+
+ if (host->link_mode_masks_nwords > 0) {
+ for (i = 0; i < host->link_mode_masks_nwords * 3; ++i) {
+ host->link_mode_masks[i] = tswap32(target->link_mode_masks[i]);
+ }
+ }
+}
+
+static int target_ethtool_link_settings_size(const void *src)
+{
+ const struct ethtool_link_settings *target = src;
+ if (target->link_mode_masks_nwords > 0) {
+ return sizeof(struct ethtool_link_settings) +
+ 3 * target->link_mode_masks_nwords *
+ sizeof(target->link_mode_masks[0]);
+ } else {
+ return sizeof(struct ethtool_link_settings);
+ }
+}
+
+static int host_ethtool_link_settings_size(const void *src)
+{
+ const struct ethtool_link_settings *host = src;
+ if (host->link_mode_masks_nwords > 0) {
+ return sizeof(struct ethtool_link_settings) +
+ 3 * host->link_mode_masks_nwords *
+ sizeof(host->link_mode_masks[0]);
+ } else {
+ return sizeof(struct ethtool_link_settings);
+ }
+}
+
+const StructEntry struct_ethtool_link_settings_def = {
+ .convert = {
+ host_to_target_ethtool_link_settings,
+ target_to_host_ethtool_link_settings
+ },
+ .thunk_size = {
+ target_ethtool_link_settings_size, host_ethtool_link_settings_size },
+ .size = {
+ sizeof(struct ethtool_link_settings),
+ sizeof(struct ethtool_link_settings) },
+ .align = {
+ __alignof__(struct ethtool_link_settings),
+ __alignof__(struct ethtool_link_settings) },
+};
+
+/* struct ethtool_per_queue_op {
+ * __u32 cmd;
+ * __u32 sub_command;
+ * __u32 queue_mask[__KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32)];
+ * char data[];
+ * };
+ *
+ * `queue_mask` are a series of bitmasks of the queues. `data` is a complete
+ * command structure for each of the queues addressed.
+ *
+ * When `cmd` is `ETHTOOL_PERQUEUE` and `sub_command` is `ETHTOOL_GCOALESCE` or
+ * `ETHTOOL_SCOALESCE`, the command structure is `struct ethtool_coalesce`.
+ */
+static void host_to_target_ethtool_per_queue_op(void *dst, const void *src)
+{
+ static const argtype ethtool_coalesce_argtype[] = {
+ MK_STRUCT(STRUCT_ethtool_coalesce), TYPE_NULL };
+ int i, queue_count;
+ struct ethtool_per_queue_op *target = dst;
+ const struct ethtool_per_queue_op *host = src;
+
+ target->cmd = tswap32(host->cmd);
+ target->sub_command = tswap32(host->sub_command);
+
+ queue_count = 0;
+ for (i = 0; i < __KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32); ++i) {
+ target->queue_mask[i] = tswap32(host->queue_mask[i]);
+ queue_count += ctpop32(host->queue_mask[i]);
+ }
+
+ if (host->cmd != ETHTOOL_PERQUEUE ||
+ (host->sub_command != ETHTOOL_GCOALESCE &&
+ host->sub_command != ETHTOOL_SCOALESCE)) {
+ fprintf(stderr,
+ "Unknown command 0x%x sub_command 0x%x for "
+ "ethtool_per_queue_op, unable to convert the `data` field "
+ "(host-to-target)\n",
+ host->cmd, host->sub_command);
+ return;
+ }
+
+ for (i = 0; i < queue_count; ++i) {
+ thunk_convert(target->data + i * sizeof(struct ethtool_coalesce),
+ host->data + i * sizeof(struct ethtool_coalesce),
+ ethtool_coalesce_argtype, THUNK_TARGET);
+ }
+}
+
+static void target_to_host_ethtool_per_queue_op(void *dst, const void *src)
+{
+ static const argtype ethtool_coalesce_argtype[] = {
+ MK_STRUCT(STRUCT_ethtool_coalesce), TYPE_NULL };
+ int i, queue_count;
+ struct ethtool_per_queue_op *host = dst;
+ const struct ethtool_per_queue_op *target = src;
+
+ host->cmd = tswap32(target->cmd);
+ host->sub_command = tswap32(target->sub_command);
+
+ queue_count = 0;
+ for (i = 0; i < __KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32); ++i) {
+ host->queue_mask[i] = tswap32(target->queue_mask[i]);
+ queue_count += ctpop32(host->queue_mask[i]);
+ }
+
+ if (host->cmd != ETHTOOL_PERQUEUE ||
+ (host->sub_command != ETHTOOL_GCOALESCE &&
+ host->sub_command != ETHTOOL_SCOALESCE)) {
+ fprintf(stderr,
+ "Unknown command 0x%x sub_command 0x%x for "
+ "ethtool_per_queue_op, unable to convert the `data` field "
+ "(target-to-host)\n",
+ host->cmd, host->sub_command);
+ return;
+ }
+
+ for (i = 0; i < queue_count; ++i) {
+ thunk_convert(host->data + i * sizeof(struct ethtool_coalesce),
+ target->data + i * sizeof(struct ethtool_coalesce),
+ ethtool_coalesce_argtype, THUNK_HOST);
+ }
+}
+
+static int target_ethtool_per_queue_op_size(const void *src)
+{
+ int i, queue_count;
+ const struct ethtool_per_queue_op *target = src;
+
+ if (tswap32(target->cmd) != ETHTOOL_PERQUEUE ||
+ (tswap32(target->sub_command) != ETHTOOL_GCOALESCE &&
+ tswap32(target->sub_command) != ETHTOOL_SCOALESCE)) {
+ fprintf(stderr,
+ "Unknown command 0x%x sub_command 0x%x for "
+ "ethtool_per_queue_op, unable to compute the size of the "
+ "`data` field (target)\n",
+ tswap32(target->cmd), tswap32(target->sub_command));
+ return sizeof(struct ethtool_per_queue_op);
+ }
+
+ queue_count = 0;
+ for (i = 0; i < __KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32); ++i) {
+ queue_count += ctpop32(target->queue_mask[i]);
+ }
+ return sizeof(struct ethtool_per_queue_op) +
+ queue_count * sizeof(struct ethtool_coalesce);
+}
+
+static int host_ethtool_per_queue_op_size(const void *src)
+{
+ int i, queue_count;
+ const struct ethtool_per_queue_op *host = src;
+
+ if (host->cmd != ETHTOOL_PERQUEUE ||
+ (host->sub_command != ETHTOOL_GCOALESCE &&
+ host->sub_command != ETHTOOL_SCOALESCE)) {
+ fprintf(stderr,
+ "Unknown command 0x%x sub_command 0x%x for "
+ "ethtool_per_queue_op, unable to compute the size of the "
+ "`data` field (host)\n",
+ host->cmd, host->sub_command);
+ return sizeof(struct ethtool_per_queue_op);
+ }
+
+ queue_count = 0;
+ for (i = 0; i < __KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32); ++i) {
+ queue_count += ctpop32(host->queue_mask[i]);
+ }
+ return sizeof(struct ethtool_per_queue_op) +
+ queue_count * sizeof(struct ethtool_coalesce);
+}
+
+const StructEntry struct_ethtool_per_queue_op_def = {
+ .convert = {
+ host_to_target_ethtool_per_queue_op,
+ target_to_host_ethtool_per_queue_op
+ },
+ .thunk_size = {
+ target_ethtool_per_queue_op_size, host_ethtool_per_queue_op_size },
+ .size = {
+ sizeof(struct ethtool_per_queue_op),
+ sizeof(struct ethtool_per_queue_op) },
+ .align = {
+ __alignof__(struct ethtool_per_queue_op),
+ __alignof__(struct ethtool_per_queue_op) },
+};
+
+#define safe_dev_ethtool(fd,...) \
+ safe_syscall(__NR_ioctl, (fd), SIOCETHTOOL, __VA_ARGS__)
+
+typedef struct EthtoolEntry EthtoolEntry;
+
+typedef abi_long do_ethtool_fn(const EthtoolEntry *ee, uint8_t *buf_temp,
+ int fd, struct ifreq *host_ifreq);
+
+struct EthtoolEntry {
+ uint32_t cmd;
+ int access;
+ do_ethtool_fn *do_ethtool;
+ const argtype arg_type[3];
+};
+
+#define ETHT_R 0x0001
+#define ETHT_W 0x0002
+#define ETHT_RW (ETHT_R | ETHT_W)
+
+static do_ethtool_fn do_ethtool_get_rxfh;
+
+static EthtoolEntry ethtool_entries[] = {
+#define ETHTOOL(cmd, access, ...) \
+ { cmd, access, 0, { __VA_ARGS__ } },
+#define ETHTOOL_SPECIAL(cmd, access, dofn, ...) \
+ { cmd, access, dofn, { __VA_ARGS__ } },
+#include "ethtool_entries.h"
+#undef ETHTOOL
+#undef ETHTOOL_SPECIAL
+ { 0, 0 },
+};
+
+/* ETHTOOL_GRSSH has two modes of operations: querying the sizes of the indir
+ * and key as well as actually querying the indir and key. When either
+ * `indir_size` or `key_size` is zero, the size of the corresponding entry is
+ * retrieved and updated into the `ethtool_rxfh` struct. When either of them is
+ * non-zero, the actually indir or key is written to `rss_config`.
+ *
+ * This causes a problem for the generic framework which converts between host
+ * and target structures without the context. When the convertion function sees
+ * an `ethtool_rxfh` struct with non-zero `indir_size` or `key_size`, it has to
+ * assume that there are entries in `rss_config` and needs to convert them.
+ * Unfortunately, when converting the returned `ethtool_rxfh` struct from host
+ * to target after an ETHTOOL_GRSSH call with the first mode, the `indir_size`
+ * and `key_size` fields are populated but there is no actual data to be
+ * converted. More importantly, user programs would not have prepared enough
+ * memory for the convertion to take place safely.
+ *
+ * ETHTOOL_GRSSH thus needs a special implementation which is aware of the two
+ * modes of operations and converts the structure accordingly.
+ */
+abi_long do_ethtool_get_rxfh(const EthtoolEntry *ee, uint8_t *buf_temp,
+ int fd, struct ifreq *host_ifreq)
+{
+ const argtype *arg_type = ee->arg_type;
+ const abi_long ifreq_data = (abi_long)(unsigned long)host_ifreq->ifr_data;
+ struct ethtool_rxfh *rxfh = (struct ethtool_rxfh *)buf_temp;
+ uint32_t user_indir_size, user_key_size;
+ abi_long ret;
+ void *argptr;
+
+ assert(arg_type[0] == TYPE_PTR);
+ assert(ee->access == IOC_RW);
+ arg_type++;
+
+ /* As of Linux kernel v5.8-rc4, ETHTOOL_GRSSH calls never read the
+ * `rss_config` part. Converting only the "header" part suffices.
+ */
+ argptr = lock_user(VERIFY_READ, ifreq_data, sizeof(*rxfh), 1);
+ if (!argptr)
+ return -TARGET_EFAULT;
+ convert_ethtool_rxfh_header(rxfh, argptr);
+ unlock_user(argptr, ifreq_data, sizeof(*rxfh));
+
+ if (rxfh->cmd != ETHTOOL_GRSSH)
+ return -TARGET_EINVAL;
+ user_indir_size = rxfh->indir_size;
+ user_key_size = rxfh->key_size;
+
+ host_ifreq->ifr_data = (void *)rxfh;
+ ret = get_errno(safe_dev_ethtool(fd, host_ifreq));
+
+ /* When a user program supplies `indir_size` or `key_size` but does not
+ * match what the kernel has, the syscall returns EINVAL but the structure
+ * is already updated. Mimicking it here.
+ */
+ argptr = lock_user(VERIFY_WRITE, ifreq_data, sizeof(*rxfh), 0);
+ if (!argptr)
+ return -TARGET_EFAULT;
+ convert_ethtool_rxfh_header(argptr, rxfh);
+ unlock_user(argptr, ifreq_data, 0);
+
+ if (is_error(ret))
+ return ret;
+
+ if (user_indir_size > 0 || user_key_size > 0) {
+ const int rss_config_size =
+ user_indir_size * sizeof(rxfh->rss_config[0]) + user_key_size;
+ argptr = lock_user(VERIFY_WRITE, ifreq_data + sizeof(*rxfh),
+ rss_config_size, 0);
+ if (!argptr)
+ return -TARGET_EFAULT;
+ convert_ethtool_rxfh_rss_config(argptr, rxfh->rss_config,
+ user_indir_size, user_key_size);
+ unlock_user(argptr, ifreq_data + sizeof(*rxfh), rss_config_size);
+ }
+ return ret;
+}
+
+/* Calculates the size of the data type represented by `type_ptr` with
+ * `guest_addr` being the underlying memory. Since `type_ptr` may contain
+ * flexible arrays, we need access to the underlying memory to determine their
+ * sizes.
+ */
+static int thunk_size(abi_long guest_addr, const argtype *type_ptr)
+{
+ /* lock_user based on `thunk_type_size` then call `thunk_type_size_with_src`
+ * on it.
+ */
+ void *src;
+ int type_size = thunk_type_size(type_ptr, /*is_host=*/0);
+ if (!thunk_type_has_flexible_array(type_ptr)) {
+ return type_size;
+ }
+
+ src = lock_user(VERIFY_READ, guest_addr, type_size, 0);
+ type_size = thunk_type_size_with_src(src, type_ptr, /*is_host=*/0);
+ unlock_user(src, guest_addr, 0);
+
+ return type_size;
+}
+
+abi_long dev_ethtool(int fd, uint8_t *buf_temp)
+{
+ uint32_t *cmd;
+ uint32_t host_cmd;
+ const EthtoolEntry *ee;
+ const argtype *arg_type;
+ abi_long ret;
+ int target_size;
+ void *argptr;
+
+ /* Make a copy of `host_ifreq` because we are going to reuse `buf_temp` and
+ * overwrite it. Further, we will overwrite `host_ifreq.ifreq_data`, so
+ * keep a copy in `ifreq_data`.
+ */
+ struct ifreq host_ifreq = *(struct ifreq *)(unsigned long)buf_temp;
+ const abi_long ifreq_data = (abi_long)(unsigned long)host_ifreq.ifr_data;
+
+ cmd = (uint32_t *)lock_user(VERIFY_READ, ifreq_data, sizeof(uint32_t), 0);
+ host_cmd = tswap32(*cmd);
+ unlock_user(cmd, ifreq_data, 0);
+
+ ee = ethtool_entries;
+ for(;;) {
+ if (ee->cmd == 0) {
+ qemu_log_mask(LOG_UNIMP, "Unsupported ethtool cmd=0x%04lx\n",
+ (long)host_cmd);
+ return -TARGET_ENOSYS;
+ }
+ if (ee->cmd == host_cmd) {
+ break;
+ }
+ ee++;
+ }
+ if (ee->do_ethtool) {
+ return ee->do_ethtool(ee, buf_temp, fd, &host_ifreq);
+ }
+
+ host_ifreq.ifr_data = buf_temp;
+ /* Even for ETHT_R, cmd still needs to be copied. */
+ *(uint32_t *)buf_temp = host_cmd;
+
+ arg_type = ee->arg_type;
+ switch (arg_type[0]) {
+ case TYPE_NULL:
+ /* no argument other than cmd */
+ ret = get_errno(safe_dev_ethtool(fd, &host_ifreq));
+ break;
+ case TYPE_PTR:
+ arg_type++;
+ target_size = thunk_size(ifreq_data, arg_type);
+ switch (ee->access) {
+ case ETHT_R:
+ ret = get_errno(safe_dev_ethtool(fd, &host_ifreq));
+ if (!is_error(ret)) {
+ argptr = lock_user(VERIFY_WRITE, ifreq_data, target_size, 0);
+ if (!argptr) {
+ return -TARGET_EFAULT;
+ }
+ thunk_convert(argptr, buf_temp, arg_type, THUNK_TARGET);
+ unlock_user(argptr, ifreq_data, target_size);
+ }
+ break;
+ case ETHT_W:
+ argptr = lock_user(VERIFY_READ, ifreq_data, target_size, 1);
+ if (!argptr) {
+ return -TARGET_EFAULT;
+ }
+ thunk_convert(buf_temp, argptr, arg_type, THUNK_HOST);
+ unlock_user(argptr, ifreq_data, 0);
+ ret = get_errno(safe_dev_ethtool(fd, &host_ifreq));
+ break;
+ default:
+ case ETHT_RW:
+ argptr = lock_user(VERIFY_READ, ifreq_data, target_size, 1);
+ if (!argptr)
+ return -TARGET_EFAULT;
+ thunk_convert(buf_temp, argptr, arg_type, THUNK_HOST);
+ unlock_user(argptr, ifreq_data, 0);
+ ret = get_errno(safe_dev_ethtool(fd, &host_ifreq));
+ if (!is_error(ret)) {
+ argptr = lock_user(VERIFY_WRITE, ifreq_data, target_size, 0);
+ if (!argptr) {
+ return -TARGET_EFAULT;
+ }
+ thunk_convert(argptr, buf_temp, arg_type, THUNK_TARGET);
+ unlock_user(argptr, ifreq_data, target_size);
+ }
+ break;
+ }
+ break;
+ default:
+ qemu_log_mask(LOG_UNIMP,
+ "Unsupported ethtool type: cmd=0x%04lx type=%d\n",
+ (long)host_cmd, arg_type[0]);
+ ret = -TARGET_ENOSYS;
+ break;
+ }
+ return ret;
+}
diff --git a/linux-user/ethtool.h b/linux-user/ethtool.h
new file mode 100644
index 0000000000..24de5dccb3
--- /dev/null
+++ b/linux-user/ethtool.h
@@ -0,0 +1,19 @@
+#ifndef ETHTOOL_H
+#define ETHTOOL_H
+
+#include <linux/if.h>
+#include "qemu.h"
+
+extern const StructEntry struct_ethtool_rxnfc_get_set_rxfh_def;
+extern const StructEntry struct_ethtool_sset_info_def;
+extern const StructEntry struct_ethtool_rxfh_def;
+extern const StructEntry struct_ethtool_link_settings_def;
+extern const StructEntry struct_ethtool_per_queue_op_def;
+
+/* Takes the file descriptor and the buffer for temporarily storing data read
+ * from / to be written to guest memory. `buf_temp` must now contain the host
+ * representation of `struct ifreq`.
+ */
+abi_long dev_ethtool(int fd, uint8_t *buf_temp);
+
+#endif /* ETHTOOL_H */
diff --git a/linux-user/ethtool_entries.h b/linux-user/ethtool_entries.h
new file mode 100644
index 0000000000..14f4e80a21
--- /dev/null
+++ b/linux-user/ethtool_entries.h
@@ -0,0 +1,107 @@
+ ETHTOOL(ETHTOOL_GSET, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_cmd)))
+ ETHTOOL(ETHTOOL_SSET, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_cmd)))
+ ETHTOOL(ETHTOOL_GDRVINFO, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_drvinfo)))
+ ETHTOOL(ETHTOOL_GREGS, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_regs)))
+ ETHTOOL(ETHTOOL_GWOL, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_wolinfo)))
+ ETHTOOL(ETHTOOL_SWOL, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_wolinfo)))
+ ETHTOOL(ETHTOOL_GMSGLVL, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_SMSGLVL, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GEEE, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_eee)))
+ ETHTOOL(ETHTOOL_SEEE, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_eee)))
+ ETHTOOL(ETHTOOL_NWAY_RST, 0, TYPE_NULL)
+ ETHTOOL(ETHTOOL_GLINK, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GEEPROM, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_eeprom)))
+ ETHTOOL(ETHTOOL_SEEPROM, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_eeprom)))
+ ETHTOOL(ETHTOOL_GCOALESCE, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_coalesce)))
+ ETHTOOL(ETHTOOL_SCOALESCE, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_coalesce)))
+ ETHTOOL(ETHTOOL_GRINGPARAM, ETHT_R,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_ringparam)))
+ ETHTOOL(ETHTOOL_SRINGPARAM, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_ringparam)))
+ ETHTOOL(ETHTOOL_GPAUSEPARAM, ETHT_R,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_pauseparam)))
+ ETHTOOL(ETHTOOL_SPAUSEPARAM, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_pauseparam)))
+ ETHTOOL(ETHTOOL_TEST, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_test)))
+ ETHTOOL(ETHTOOL_GSTRINGS, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_gstrings)))
+ ETHTOOL(ETHTOOL_PHYS_ID, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GSTATS, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_stats)))
+ ETHTOOL(ETHTOOL_GPERMADDR, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_perm_addr)))
+ ETHTOOL(ETHTOOL_GFLAGS, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_SFLAGS, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GPFLAGS, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_SPFLAGS, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GRXFH, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_get_set_rxfh)))
+ ETHTOOL(ETHTOOL_GRXRINGS, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_rss_context)))
+ ETHTOOL(ETHTOOL_GRXCLSRLCNT, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_rule_cnt)))
+ ETHTOOL(ETHTOOL_GRXCLSRULE, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_rss_context)))
+ ETHTOOL(ETHTOOL_GRXCLSRLALL, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_rule_locs)))
+ ETHTOOL(ETHTOOL_SRXFH, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_get_set_rxfh)))
+ ETHTOOL(ETHTOOL_SRXCLSRLDEL, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_rss_context)))
+ ETHTOOL(ETHTOOL_SRXCLSRLINS, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxnfc_rss_context)))
+ ETHTOOL(ETHTOOL_FLASHDEV, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_flash)))
+ ETHTOOL(ETHTOOL_RESET, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GSSET_INFO, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_sset_info)))
+ ETHTOOL(ETHTOOL_GRXFHINDIR, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxfh_indir)))
+ ETHTOOL(ETHTOOL_SRXFHINDIR, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxfh_indir)))
+ ETHTOOL_SPECIAL(ETHTOOL_GRSSH, ETHT_RW, do_ethtool_get_rxfh,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxfh)))
+ ETHTOOL(ETHTOOL_SRSSH, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_rxfh)))
+ ETHTOOL(ETHTOOL_GFEATURES, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_gfeatures)))
+ ETHTOOL(ETHTOOL_SFEATURES, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_sfeatures)))
+ ETHTOOL(ETHTOOL_GTXCSUM, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GRXCSUM, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GSG, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GTSO, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GGSO, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GGRO, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_STXCSUM, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_SRXCSUM, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_SSG, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_STSO, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_SGSO, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_SGRO, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_value)))
+ ETHTOOL(ETHTOOL_GCHANNELS, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_channels)))
+ ETHTOOL(ETHTOOL_SCHANNELS, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_channels)))
+ ETHTOOL(ETHTOOL_SET_DUMP, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_dump_no_data)))
+ ETHTOOL(ETHTOOL_GET_DUMP_FLAG, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_dump_no_data)))
+ ETHTOOL(ETHTOOL_GET_DUMP_DATA, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_dump)))
+ ETHTOOL(ETHTOOL_GET_TS_INFO, ETHT_R,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_ts_info)))
+ ETHTOOL(ETHTOOL_GMODULEINFO, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_modinfo)))
+ ETHTOOL(ETHTOOL_GMODULEEEPROM, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_eeprom)))
+ ETHTOOL(ETHTOOL_GTUNABLE, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_tunable)))
+ ETHTOOL(ETHTOOL_STUNABLE, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_tunable)))
+ ETHTOOL(ETHTOOL_GPHYSTATS, ETHT_RW, MK_PTR(MK_STRUCT(STRUCT_ethtool_stats)))
+ ETHTOOL(ETHTOOL_PERQUEUE, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_per_queue_op)))
+ ETHTOOL(ETHTOOL_GLINKSETTINGS, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_link_settings)))
+ ETHTOOL(ETHTOOL_SLINKSETTINGS, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_link_settings)))
+ ETHTOOL(ETHTOOL_PHY_GTUNABLE, ETHT_RW,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_tunable)))
+ ETHTOOL(ETHTOOL_PHY_STUNABLE, ETHT_W,
+ MK_PTR(MK_STRUCT(STRUCT_ethtool_tunable)))
+ ETHTOOL(ETHTOOL_GFECPARAM, ETHT_R, MK_PTR(MK_STRUCT(STRUCT_ethtool_fecparam)))
+ ETHTOOL(ETHTOOL_GFECPARAM, ETHT_W, MK_PTR(MK_STRUCT(STRUCT_ethtool_fecparam)))
diff --git a/linux-user/ioctls.h b/linux-user/ioctls.h
index 0713ae1311..fd6ac963ec 100644
--- a/linux-user/ioctls.h
+++ b/linux-user/ioctls.h
@@ -238,6 +238,8 @@
IOCTL(SIOCSIFHWADDR, IOC_W, MK_PTR(MK_STRUCT(STRUCT_sockaddr_ifreq)))
IOCTL(SIOCGIFTXQLEN, IOC_W | IOC_R, MK_PTR(MK_STRUCT(STRUCT_sockaddr_ifreq)))
IOCTL(SIOCSIFTXQLEN, IOC_W, MK_PTR(MK_STRUCT(STRUCT_sockaddr_ifreq)))
+ IOCTL_SPECIAL(SIOCETHTOOL, IOC_W | IOC_R, do_ioctl_ethtool,
+ MK_PTR(MK_STRUCT(STRUCT_ptr_ifreq)))
IOCTL(SIOCGIFMETRIC, IOC_W | IOC_R, MK_PTR(MK_STRUCT(STRUCT_int_ifreq)))
IOCTL(SIOCSIFMETRIC, IOC_W, MK_PTR(MK_STRUCT(STRUCT_int_ifreq)))
IOCTL(SIOCGIFMTU, IOC_W | IOC_R, MK_PTR(MK_STRUCT(STRUCT_int_ifreq)))
diff --git a/linux-user/qemu.h b/linux-user/qemu.h
index 5c964389c1..43f00681f8 100644
--- a/linux-user/qemu.h
+++ b/linux-user/qemu.h
@@ -231,6 +231,7 @@ abi_long do_syscall(void *cpu_env, int num, abi_long arg1,
abi_long arg2, abi_long arg3, abi_long arg4,
abi_long arg5, abi_long arg6, abi_long arg7,
abi_long arg8);
+abi_long get_errno(abi_long ret);
extern __thread CPUState *thread_cpu;
void cpu_loop(CPUArchState *env);
const char *target_strerror(int err);
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 483f8ce48f..bbdcc5d4db 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -127,6 +127,7 @@
#include "qapi/error.h"
#include "fd-trans.h"
#include "tcg/tcg.h"
+#include "ethtool.h"
#ifndef CLONE_IO
#define CLONE_IO 0x80000000 /* Clone io context */
@@ -683,7 +684,7 @@ static inline int target_to_host_errno(int err)
return err;
}
-static inline abi_long get_errno(abi_long ret)
+abi_long get_errno(abi_long ret)
{
if (ret == -1)
return -host_to_target_errno(errno);
@@ -4681,16 +4682,6 @@ static abi_long do_ipc(CPUArchState *cpu_env,
#endif
/* kernel structure types definitions */
-
-#define STRUCT(name, ...) STRUCT_ ## name,
-#define STRUCT_SPECIAL(name) STRUCT_ ## name,
-enum {
-#include "syscall_types.h"
-STRUCT_MAX
-};
-#undef STRUCT
-#undef STRUCT_SPECIAL
-
#define STRUCT(name, ...) static const argtype struct_ ## name ## _def[] = { __VA_ARGS__, TYPE_NULL };
#define STRUCT_SPECIAL(name)
#include "syscall_types.h"
@@ -4788,6 +4779,28 @@ static abi_long do_ioctl_fs_ioc_fiemap(const IOCTLEntry *ie, uint8_t *buf_temp,
}
#endif
+static abi_long do_ioctl_ethtool(const IOCTLEntry *ie, uint8_t *buf_temp,
+ int fd, int cmd, abi_long arg)
+{
+ const argtype *arg_type = ie->arg_type;
+ int target_size;
+ void *argptr;
+
+ assert(arg_type[0] == TYPE_PTR);
+ assert(ie->access == IOC_RW);
+
+ arg_type++;
+ target_size = thunk_type_size(arg_type, 0);
+
+ argptr = lock_user(VERIFY_READ, arg, target_size, 1);
+ if (!argptr)
+ return -TARGET_EFAULT;
+ thunk_convert(buf_temp, argptr, arg_type, THUNK_HOST);
+ unlock_user(argptr, arg, target_size);
+
+ return dev_ethtool(fd, buf_temp);
+}
+
static abi_long do_ioctl_ifconf(const IOCTLEntry *ie, uint8_t *buf_temp,
int fd, int cmd, abi_long arg)
{
diff --git a/linux-user/syscall_defs.h b/linux-user/syscall_defs.h
index 70df1a94fb..5d344d5ffa 100644
--- a/linux-user/syscall_defs.h
+++ b/linux-user/syscall_defs.h
@@ -866,6 +866,8 @@ struct target_rtc_pll_info {
#define TARGET_SIOCGIFTXQLEN 0x8942 /* Get the tx queue length */
#define TARGET_SIOCSIFTXQLEN 0x8943 /* Set the tx queue length */
+#define TARGET_SIOCETHTOOL 0x8946 /* Ethtool interface */
+
/* ARP cache control calls. */
#define TARGET_OLD_SIOCDARP 0x8950 /* old delete ARP table entry */
#define TARGET_OLD_SIOCGARP 0x8951 /* old get ARP table entry */
@@ -2776,4 +2778,14 @@ struct target_statx {
/* 0x100 */
};
+/* kernel structure types definitions */
+#define STRUCT(name, ...) STRUCT_ ## name,
+#define STRUCT_SPECIAL(name) STRUCT_ ## name,
+enum {
+#include "syscall_types.h"
+STRUCT_MAX
+};
+#undef STRUCT
+#undef STRUCT_SPECIAL
+
#endif
diff --git a/linux-user/syscall_types.h b/linux-user/syscall_types.h
index 3f1f033464..e920086e66 100644
--- a/linux-user/syscall_types.h
+++ b/linux-user/syscall_types.h
@@ -1,3 +1,4 @@
+
STRUCT_SPECIAL(termios)
STRUCT(winsize,
@@ -464,3 +465,279 @@ STRUCT(usbdevfs_disconnect_claim,
TYPE_INT, /* flags */
MK_ARRAY(TYPE_CHAR, USBDEVFS_MAXDRIVERNAME + 1)) /* driver */
#endif /* CONFIG_USBFS */
+
+/* ethtool ioctls */
+STRUCT(ethtool_cmd,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* supported */
+ TYPE_INT, /* advertising */
+ TYPE_SHORT, /* speed */
+ TYPE_CHAR, /* duplex */
+ TYPE_CHAR, /* port */
+ TYPE_CHAR, /* phy_address */
+ TYPE_CHAR, /* transceiver */
+ TYPE_CHAR, /* autoneg */
+ TYPE_CHAR, /* mdio_support */
+ TYPE_INT, /* maxtxpkt */
+ TYPE_INT, /* maxrxpkt */
+ TYPE_SHORT, /* speed_hi */
+ TYPE_CHAR, /* eth_tp_mdix */
+ TYPE_CHAR, /* eth_tp_mdix_ctrl */
+ TYPE_INT, /* lp_advertising */
+ MK_ARRAY(TYPE_INT, 2)) /* reserved */
+
+STRUCT(ethtool_drvinfo,
+ TYPE_INT, /* cmd */
+ MK_ARRAY(TYPE_CHAR, 32), /* driver */
+ MK_ARRAY(TYPE_CHAR, 32), /* version */
+ MK_ARRAY(TYPE_CHAR, 32), /* fw_version[ETHTOOL_FWVERS_LEN] */
+ MK_ARRAY(TYPE_CHAR, 32), /* bus_info[ETHTOOL_BUSINFO_LEN] */
+ MK_ARRAY(TYPE_CHAR, 32), /* erom_version[ETHTOOL_EROMVERS_LEN] */
+ MK_ARRAY(TYPE_CHAR, 12), /* reserved2 */
+ TYPE_INT, /* n_priv_flags */
+ TYPE_INT, /* n_stats */
+ TYPE_INT, /* testinfo_len */
+ TYPE_INT, /* eedump_len */
+ TYPE_INT) /* regdump_len */
+
+STRUCT(ethtool_regs,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* version */
+ TYPE_INT, /* len */
+ MK_FLEXIBLE_ARRAY(TYPE_CHAR, 2)) /* data[0]: len */
+
+STRUCT(ethtool_wolinfo,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* supported */
+ TYPE_INT, /* wolopts */
+ MK_ARRAY(TYPE_CHAR, 6)) /* sopass[SOPASS_MAX] */
+
+STRUCT(ethtool_value,
+ TYPE_INT, /* cmd */
+ TYPE_INT) /* data */
+
+STRUCT(ethtool_eee,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* supported */
+ TYPE_INT, /* advertised */
+ TYPE_INT, /* lp_advertised */
+ TYPE_INT, /* eee_active */
+ TYPE_INT, /* eee_enabled */
+ TYPE_INT, /* tx_lpi_enabled */
+ TYPE_INT, /* tx_lpi_timer */
+ MK_ARRAY(TYPE_INT, 2)) /* reserved */
+
+STRUCT(ethtool_eeprom,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* magic */
+ TYPE_INT, /* offset */
+ TYPE_INT, /* len */
+ MK_FLEXIBLE_ARRAY(TYPE_CHAR, 3)) /* data[0]: len */
+
+STRUCT(ethtool_coalesce,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* rx_coalesce_usecs */
+ TYPE_INT, /* rx_max_coalesced_frames */
+ TYPE_INT, /* rx_coalesce_usecs_irq */
+ TYPE_INT, /* rx_max_coalesced_frames_irq */
+ TYPE_INT, /* tx_coalesce_usecs */
+ TYPE_INT, /* tx_max_coalesced_frames */
+ TYPE_INT, /* tx_coalesce_usecs_irq */
+ TYPE_INT, /* tx_max_coalesced_frames_irq */
+ TYPE_INT, /* stats_block_coalesce_usecs */
+ TYPE_INT, /* use_adaptive_rx_coalesce */
+ TYPE_INT, /* use_adaptive_tx_coalesce */
+ TYPE_INT, /* pkt_rate_low */
+ TYPE_INT, /* rx_coalesce_usecs_low */
+ TYPE_INT, /* rx_max_coalesced_frames_low */
+ TYPE_INT, /* tx_coalesce_usecs_low */
+ TYPE_INT, /* tx_max_coalesced_frames_low */
+ TYPE_INT, /* pkt_rate_high */
+ TYPE_INT, /* rx_coalesce_usecs_high */
+ TYPE_INT, /* rx_max_coalesced_frames_high */
+ TYPE_INT, /* tx_coalesce_usecs_high */
+ TYPE_INT, /* tx_max_coalesced_frames_high */
+ TYPE_INT) /* rate_sample_interval */
+
+STRUCT(ethtool_ringparam,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* rx_max_pending */
+ TYPE_INT, /* rx_mini_max_pending */
+ TYPE_INT, /* rx_jumbo_max_pending */
+ TYPE_INT, /* tx_max_pending */
+ TYPE_INT, /* rx_pending */
+ TYPE_INT, /* rx_mini_pending */
+ TYPE_INT, /* rx_jumbo_pending */
+ TYPE_INT) /* tx_pending */
+
+STRUCT(ethtool_pauseparam,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* autoneg */
+ TYPE_INT, /* rx_pause */
+ TYPE_INT) /* tx_pause */
+
+STRUCT(ethtool_test,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* flags */
+ TYPE_INT, /* reserved */
+ TYPE_INT, /* len */
+ MK_FLEXIBLE_ARRAY(TYPE_LONGLONG, 3)) /* data[0]: len */
+
+STRUCT(ethtool_gstrings,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* string_set */
+ TYPE_INT, /* len */
+ MK_FLEXIBLE_ARRAY(MK_ARRAY(TYPE_CHAR, 32), 2))
+ /* data[0]: len * ETH_GSTRING_LEN */
+
+STRUCT(ethtool_stats,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* n_stats */
+ MK_FLEXIBLE_ARRAY(TYPE_LONGLONG, 1)) /* data[0]: n_stats */
+
+STRUCT(ethtool_perm_addr,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* size */
+ MK_FLEXIBLE_ARRAY(TYPE_CHAR, 1)) /* data[0]: size */
+
+STRUCT(ethtool_flow_ext,
+ MK_ARRAY(TYPE_CHAR, 2), /* padding */
+ MK_ARRAY(TYPE_CHAR, 6), /* h_dest[ETH_ALEN] */
+ MK_ARRAY(TYPE_CHAR, 2), /* __be16 vlan_etype */
+ MK_ARRAY(TYPE_CHAR, 2), /* __be16 vlan_tci */
+ MK_ARRAY(TYPE_CHAR, 8)) /* __be32 data[2] */
+
+/* Union ethtool_flow_union contains alternatives that are either struct that
+ * only uses __be* types or char/__u8, or "__u8 hdata[52]". We can treat it as
+ * byte array in all cases.
+ */
+STRUCT(ethtool_rx_flow_spec,
+ TYPE_INT, /* flow_type */
+ MK_ARRAY(TYPE_CHAR, 52), /* union ethtool_flow_union h_u */
+ MK_STRUCT(STRUCT_ethtool_flow_ext), /* h_ext */
+ MK_ARRAY(TYPE_CHAR, 52), /* union ethtool_flow_union m_u */
+ MK_STRUCT(STRUCT_ethtool_flow_ext), /* m_ext */
+ TYPE_LONGLONG, /* ring_cookie */
+ TYPE_INT) /* location */
+
+STRUCT(ethtool_rxnfc_rss_context,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* flow_type */
+ TYPE_LONGLONG, /* data */
+ MK_STRUCT(STRUCT_ethtool_rx_flow_spec), /* fs */
+ TYPE_INT) /* rss_context */
+
+STRUCT(ethtool_rxnfc_rule_cnt,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* flow_type */
+ TYPE_LONGLONG, /* data */
+ MK_STRUCT(STRUCT_ethtool_rx_flow_spec), /* fs */
+ TYPE_INT) /* rss_cnt */
+
+STRUCT(ethtool_rxnfc_rule_locs,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* flow_type */
+ TYPE_LONGLONG, /* data */
+ MK_STRUCT(STRUCT_ethtool_rx_flow_spec), /* fs */
+ TYPE_INT, /* rss_cnt */
+ MK_FLEXIBLE_ARRAY(TYPE_INT, 4)) /* rule_locs[0]: rss_cnt */
+
+/* For ETHTOOL_{G,S}RXFH, originally only the first three fields are defined,
+ * but with certain options, more fields are used.
+ */
+STRUCT_SPECIAL(ethtool_rxnfc_get_set_rxfh)
+
+STRUCT(ethtool_flash,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* region */
+ MK_ARRAY(TYPE_CHAR, 128)) /* data[ETHTOOL_FLASH_MAX_FILENAME] */
+
+STRUCT_SPECIAL(ethtool_sset_info)
+
+STRUCT(ethtool_rxfh_indir,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* size */
+ MK_FLEXIBLE_ARRAY(TYPE_INT, 1)) /* ring_index[0]: size */
+
+STRUCT_SPECIAL(ethtool_rxfh)
+
+STRUCT(ethtool_get_features_block,
+ TYPE_INT, /* available */
+ TYPE_INT, /* requested */
+ TYPE_INT, /* active */
+ TYPE_INT) /* never_changed */
+
+STRUCT(ethtool_gfeatures,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* size */
+ MK_FLEXIBLE_ARRAY(MK_STRUCT(STRUCT_ethtool_get_features_block), 1))
+ /* features[0]: size */
+
+STRUCT(ethtool_set_features_block,
+ TYPE_INT, /* valid */
+ TYPE_INT) /* requested */
+
+STRUCT(ethtool_sfeatures,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* size */
+ MK_FLEXIBLE_ARRAY(MK_STRUCT(STRUCT_ethtool_set_features_block), 1))
+ /* features[0]: size */
+
+STRUCT(ethtool_channels,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* max_rx */
+ TYPE_INT, /* max_tx */
+ TYPE_INT, /* max_other */
+ TYPE_INT, /* max_combined */
+ TYPE_INT, /* rx_count */
+ TYPE_INT, /* tx_count */
+ TYPE_INT, /* other_count */
+ TYPE_INT) /* combined_count */
+
+/* For ETHTOOL_SET_DUMP and ETHTOOL_GET_DUMP_FLAG, the flexible array `data` is
+ * not used.
+ */
+STRUCT(ethtool_dump_no_data,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* version */
+ TYPE_INT, /* flag */
+ TYPE_INT) /* len */
+
+STRUCT(ethtool_dump,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* version */
+ TYPE_INT, /* flag */
+ TYPE_INT, /* len */
+ MK_FLEXIBLE_ARRAY(TYPE_CHAR, 3)) /* data[0]: len */
+
+STRUCT(ethtool_ts_info,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* so_timestamping */
+ TYPE_INT, /* phc_index */
+ TYPE_INT, /* tx_types */
+ MK_ARRAY(TYPE_INT, 3), /* tx_reserved */
+ TYPE_INT, /* rx_filters */
+ MK_ARRAY(TYPE_INT, 3)) /* rx_reserved */
+
+STRUCT(ethtool_modinfo,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* type */
+ TYPE_INT, /* eeprom_len */
+ MK_ARRAY(TYPE_INT, 8)) /* reserved */
+
+STRUCT(ethtool_tunable,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* id */
+ TYPE_INT, /* type_id */
+ TYPE_INT, /* len */
+ MK_FLEXIBLE_ARRAY(TYPE_PTRVOID, 3)) /* data[0]: len */
+
+STRUCT_SPECIAL(ethtool_link_settings)
+
+STRUCT(ethtool_fecparam,
+ TYPE_INT, /* cmd */
+ TYPE_INT, /* active_fec */
+ TYPE_INT, /* fec */
+ TYPE_INT) /* reserved */
+
+STRUCT_SPECIAL(ethtool_per_queue_op)
diff --git a/tests/tcg/multiarch/ethtool.c b/tests/tcg/multiarch/ethtool.c
new file mode 100644
index 0000000000..54b2337aa7
--- /dev/null
+++ b/tests/tcg/multiarch/ethtool.c
@@ -0,0 +1,417 @@
+#include <asm-generic/errno.h>
+#include <assert.h>
+#include <errno.h>
+#include <inttypes.h>
+#include <linux/ethtool.h>
+#include <linux/if.h>
+#include <linux/sockios.h>
+#include <netinet/in.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+
+const int number_of_entries_to_print = 10;
+const uint32_t protected_memory_pattern[] = {
+ 0xdeadc0de, 0xb0bb1e, 0xfacade, 0xfeeb1e };
+
+static void fail_with(const char *action, const char *cmd_name, int cmd, int err) {
+ if (errno == EOPNOTSUPP) {
+ printf("Unsupported operation: %s; errno = %d: %s.\n"
+ "TEST SKIPPED (%s = 0x%x).\n",
+ action, err, strerror(err), cmd_name, cmd);
+ return;
+ }
+ if (err) {
+ fprintf(stderr,
+ "Failed to %s (%s = 0x%x): errno = %d: %s\n",
+ action, cmd_name, cmd, err, strerror(err));
+ } else {
+ fprintf(stderr,
+ "Failed to %s (%s = 0x%x): no errno\n",
+ action, cmd_name, cmd);
+ }
+ exit(err);
+}
+#define FAIL(action,cmd) fail_with(action, #cmd, cmd, errno)
+
+/* `calloc_protected` and `protected_memory_changed` can be used to verify that
+ * a system call does not write pass intended memory boundary.
+ *
+ * `ptr = calloc_protected(n)` will allocate extra memory after `n` bytes and
+ * populate it with a memory pattern. The first `n` bytes are still guaranteed
+ * to be zeroed out like `calloc(1, n)`. `protected_memory_changed(ptr, n)`
+ * takes the pointer and the original size `n` and checks that the memory
+ * pattern is intact.
+ */
+uint8_t *calloc_protected(size_t struct_size)
+{
+ uint8_t *buf = (uint8_t *) calloc(
+ 1,
+ struct_size + sizeof(protected_memory_pattern));
+ memcpy(buf + struct_size, protected_memory_pattern,
+ sizeof(protected_memory_pattern));
+ return buf;
+}
+
+bool protected_memory_changed(const uint8_t *ptr, size_t struct_size)
+{
+ return memcmp(ptr + struct_size, protected_memory_pattern,
+ sizeof(protected_memory_pattern)) != 0;
+}
+
+void print_entries(const char *fmt, int len, uint32_t *entries)
+{
+ int i;
+ for (i = 0; i < len && i < number_of_entries_to_print; ++i) {
+ printf(fmt, entries[i]);
+ }
+ if (len > number_of_entries_to_print) {
+ printf(" (%d more omitted)", len - number_of_entries_to_print);
+ }
+}
+
+void basic_test(int socketfd, struct ifreq ifr)
+{
+ struct ethtool_drvinfo drvinfo;
+ drvinfo.cmd = ETHTOOL_GDRVINFO;
+ ifr.ifr_data = (void *)&drvinfo;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get driver info", ETHTOOL_GDRVINFO);
+ return;
+ }
+ printf("Driver: %s (version %s)\n", drvinfo.driver, drvinfo.version);
+}
+
+/* Test flexible array. */
+void test_get_stats(int socketfd, struct ifreq ifr, int n_stats)
+{
+ int i;
+ struct ethtool_stats *stats = (struct ethtool_stats *)calloc(
+ 1, sizeof(*stats) + sizeof(stats->data[0]) * n_stats);
+ stats->cmd = ETHTOOL_GSTATS;
+ stats->n_stats = n_stats;
+ ifr.ifr_data = (void *)stats;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get statastics", ETHTOOL_GSTATS);
+ free(stats);
+ return;
+ }
+ if (stats->n_stats != n_stats) {
+ FAIL("get consistent number of statistics", ETHTOOL_GSTATS);
+ }
+ for (i = 0; i < stats->n_stats && i < number_of_entries_to_print; ++i) {
+ printf("stats[%d] = %llu\n", i, (unsigned long long)stats->data[i]);
+ }
+ if (stats->n_stats > number_of_entries_to_print) {
+ printf("(%d more omitted)\n",
+ stats->n_stats - number_of_entries_to_print);
+ }
+ free(stats);
+}
+
+/* Test flexible array with char array as elements. */
+void test_get_strings(int socketfd, struct ifreq ifr, int n_stats)
+{
+ int i;
+ struct ethtool_gstrings *gstrings =
+ (struct ethtool_gstrings *)calloc(
+ 1, sizeof(*gstrings) + ETH_GSTRING_LEN * n_stats);
+ gstrings->cmd = ETHTOOL_GSTRINGS;
+ gstrings->string_set = ETH_SS_STATS;
+ gstrings->len = n_stats;
+ ifr.ifr_data = (void *)gstrings;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get string set", ETHTOOL_GSTRINGS);
+ free(gstrings);
+ return;
+ }
+ if (gstrings->len != n_stats) {
+ FAIL("get consistent number of statistics", ETHTOOL_GSTRINGS);
+ }
+ for (i = 0; i < gstrings->len && i < number_of_entries_to_print; ++i) {
+ printf("stat_names[%d] = %.*s\n",
+ i, ETH_GSTRING_LEN, gstrings->data + i * ETH_GSTRING_LEN);
+ }
+ if (gstrings->len > number_of_entries_to_print) {
+ printf("(%d more omitted)\n",
+ gstrings->len - number_of_entries_to_print);
+ }
+ free(gstrings);
+}
+
+/* Testing manual implementation of converting `struct ethtool_sset_info`, also
+ * info for subsequent tests.
+ */
+int test_get_sset_info(int socketfd, struct ifreq ifr)
+{
+ const int n_sset = 2;
+ int n_stats;
+ struct ethtool_sset_info *sset_info =
+ (struct ethtool_sset_info *)calloc(
+ 1, sizeof(*sset_info) + sizeof(sset_info->data[0]) * n_sset);
+ sset_info->cmd = ETHTOOL_GSSET_INFO;
+ sset_info->sset_mask = 1 << ETH_SS_TEST | 1 << ETH_SS_STATS;
+ assert(__builtin_popcount(sset_info->sset_mask) == n_sset);
+ ifr.ifr_data = (void *)sset_info;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ fail_with("get string set info", "ETHTOOL_GSSET_INFO",
+ ETHTOOL_GSSET_INFO, errno);
+ free(sset_info);
+ return 0;
+ }
+ if ((sset_info->sset_mask & (1 << ETH_SS_STATS)) == 0) {
+ puts("No stats string set info, SKIPPING dependent tests");
+ free(sset_info);
+ return 0;
+ }
+ n_stats = (sset_info->sset_mask & (1 << ETH_SS_TEST)) ?
+ sset_info->data[1] :
+ sset_info->data[0];
+ printf("n_stats = %d\n", n_stats);
+ free(sset_info);
+ return n_stats;
+}
+
+/* Test manual implementation of converting `struct ethtool_rxnfc`, focusing on
+ * the case where only the first three fields are present. (The original struct
+ * definition.)
+ */
+void test_get_rxfh(int socketfd, struct ifreq ifr)
+{
+ struct ethtool_rxnfc *rxnfc;
+ const int rxnfc_first_three_field_size =
+ sizeof(rxnfc->cmd) + sizeof(rxnfc->flow_type) + sizeof(rxnfc->data);
+ rxnfc = (struct ethtool_rxnfc *)calloc_protected(
+ rxnfc_first_three_field_size);
+ rxnfc->cmd = ETHTOOL_GRXFH;
+ rxnfc->flow_type = TCP_V4_FLOW;
+ ifr.ifr_data = (void *)rxnfc;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get RX flow classification rules", ETHTOOL_GRXFH);
+ free(rxnfc);
+ return;
+ }
+ if (protected_memory_changed((const uint8_t *)rxnfc,
+ rxnfc_first_three_field_size)) {
+ FAIL("preserve memory after the first three fields", ETHTOOL_GRXFH);
+ }
+ printf("Flow hash bitmask (flow_type = TCP v4): 0x%llx\n",
+ (unsigned long long)rxnfc->data);
+ free(rxnfc);
+}
+
+/* Test manual implementation of converting `struct ethtool_link_settings`. */
+void test_get_link_settings(int socketfd, struct ifreq ifr)
+{
+ int link_mode_masks_nwords;
+ struct ethtool_link_settings *link_settings_header =
+ (struct ethtool_link_settings *) calloc_protected(
+ sizeof(*link_settings_header));
+ link_settings_header->cmd = ETHTOOL_GLINKSETTINGS;
+ link_settings_header->link_mode_masks_nwords = 0;
+ ifr.ifr_data = (void*)link_settings_header;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get link settings mask sizes", ETHTOOL_GLINKSETTINGS);
+ free(link_settings_header);
+ return;
+ }
+ if (protected_memory_changed((const uint8_t *)link_settings_header,
+ sizeof(*link_settings_header))) {
+ FAIL("preserve link_mode_masks", ETHTOOL_GLINKSETTINGS);
+ }
+ if (link_settings_header->link_mode_masks_nwords >= 0) {
+ FAIL("complete handshake", ETHTOOL_GLINKSETTINGS);
+ }
+ link_mode_masks_nwords = -link_settings_header->link_mode_masks_nwords;
+
+ struct ethtool_link_settings *link_settings =
+ (struct ethtool_link_settings *)calloc(
+ 1,
+ sizeof(*link_settings) +
+ sizeof(link_settings_header->link_mode_masks[0]) *
+ link_mode_masks_nwords * 3);
+ link_settings->cmd = ETHTOOL_GLINKSETTINGS;
+ link_settings->link_mode_masks_nwords = link_mode_masks_nwords;
+ ifr.ifr_data = (void *)link_settings;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get link settings", ETHTOOL_GLINKSETTINGS);
+ free(link_settings_header);
+ free(link_settings);
+ return;
+ }
+ if (link_settings->link_mode_masks_nwords != link_mode_masks_nwords) {
+ FAIL("have consistent number of mode masks", ETHTOOL_GLINKSETTINGS);
+ }
+
+ printf("Link speed: %d MB\n", link_settings->speed);
+ printf("Number of link mode masks: %d\n",
+ link_settings->link_mode_masks_nwords);
+ if (link_settings->link_mode_masks_nwords > 0) {
+ printf("Supported bitmap:");
+ print_entries(" 0x%08x",
+ link_settings->link_mode_masks_nwords,
+ link_settings->link_mode_masks);
+ putchar('\n');
+
+ printf("Advertising bitmap:");
+ print_entries(" 0x%08x",
+ link_settings->link_mode_masks_nwords,
+ link_settings->link_mode_masks +
+ link_settings->link_mode_masks_nwords);
+ putchar('\n');
+
+ printf("Lp advertising bitmap:");
+ print_entries(" 0x%08x",
+ link_settings->link_mode_masks_nwords,
+ link_settings->link_mode_masks +
+ 2 * link_settings->link_mode_masks_nwords);
+ putchar('\n');
+ }
+
+ free(link_settings_header);
+ free(link_settings);
+}
+
+/* Test manual implementation of converting `struct ethtool_per_queue_op`. */
+void test_perqueue(int socketfd, struct ifreq ifr)
+{
+ const int n_queue = 2;
+ int i;
+ struct ethtool_per_queue_op *per_queue_op =
+ (struct ethtool_per_queue_op *)calloc(
+ 1,
+ sizeof(*per_queue_op) + sizeof(struct ethtool_coalesce) * n_queue);
+ per_queue_op->cmd = ETHTOOL_PERQUEUE;
+ per_queue_op->sub_command = ETHTOOL_GCOALESCE;
+ per_queue_op->queue_mask[0] = 0x3;
+ ifr.ifr_data = (void *)per_queue_op;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get coalesce per queue", ETHTOOL_PERQUEUE);
+ free(per_queue_op);
+ return;
+ }
+ for (i = 0; i < n_queue; ++i) {
+ struct ethtool_coalesce *coalesce = (struct ethtool_coalesce *)(
+ per_queue_op->data + sizeof(*coalesce) * i);
+ if (coalesce->cmd != ETHTOOL_GCOALESCE) {
+ fprintf(stderr,
+ "ETHTOOL_PERQUEUE (%d) sub_command ETHTOOL_GCOALESCE (%d) "
+ "fails to set entry %d's cmd to ETHTOOL_GCOALESCE, got %d "
+ "instead\n",
+ ETHTOOL_PERQUEUE, ETHTOOL_GCOALESCE, i,
+ coalesce->cmd);
+ exit(-1);
+ }
+ printf("rx_coalesce_usecs[%d] = %u\nrx_max_coalesced_frames[%d] = %u\n",
+ i, coalesce->rx_coalesce_usecs,
+ i, coalesce->rx_max_coalesced_frames);
+ }
+
+ free(per_queue_op);
+}
+
+/* Test manual implementation of ETHTOOL_GRSSH. */
+void test_get_rssh(int socketfd, struct ifreq ifr)
+{
+ int i;
+ struct ethtool_rxfh *rxfh_header =
+ (struct ethtool_rxfh *)calloc_protected(sizeof(*rxfh_header));
+ rxfh_header->cmd = ETHTOOL_GRSSH;
+ ifr.ifr_data = (void *)rxfh_header;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get RX flow hash indir and hash key size", ETHTOOL_GRSSH);
+ free(rxfh_header);
+ return;
+ }
+ if (protected_memory_changed((const uint8_t *)rxfh_header,
+ sizeof(*rxfh_header))) {
+ FAIL("preserve rss_config", ETHTOOL_GRSSH);
+ }
+ printf("RX flow hash indir size = %d\nRX flow hash key size = %d\n",
+ rxfh_header->indir_size, rxfh_header->key_size);
+
+ struct ethtool_rxfh *rxfh = (struct ethtool_rxfh *)calloc(
+ 1,
+ sizeof(*rxfh) + 4 * rxfh_header->indir_size + rxfh_header->key_size);
+ *rxfh = *rxfh_header;
+ ifr.ifr_data = (void*)rxfh;
+ if (ioctl(socketfd, SIOCETHTOOL, &ifr) == -1) {
+ FAIL("get RX flow hash indir and hash key", ETHTOOL_GRSSH);
+ free(rxfh_header);
+ free(rxfh);
+ return;
+ }
+
+ if (rxfh->indir_size == 0) {
+ printf("No RX flow hash indir\n");
+ } else {
+ printf("RX flow hash indir:");
+ print_entries(" 0x%08x", rxfh->indir_size, rxfh->rss_config);
+ putchar('\n');
+ }
+
+ if (rxfh->key_size == 0) {
+ printf("No RX flow hash key\n");
+ } else {
+ char *key = (char *)(rxfh->rss_config + rxfh->indir_size);
+ printf("RX flow hash key:");
+ for (i = 0; i < rxfh->key_size; ++i) {
+ if (i % 2 == 0)
+ putchar(' ');
+ printf("%02hhx", key[i]);
+ }
+ putchar('\n');
+ }
+ free(rxfh_header);
+ free(rxfh);
+}
+
+int main(int argc, char **argv)
+{
+ int socketfd, n_stats, i;
+ struct ifreq ifr;
+
+ socketfd = socket(AF_INET, SOCK_DGRAM, 0);
+ if (socketfd == -1) {
+ int err = errno;
+ fprintf(stderr,
+ "Failed to open socket: errno = %d: %s\n",
+ err, strerror(err));
+ return err;
+ }
+
+ for (i = 1;; ++i) {
+ ifr.ifr_ifindex = i;
+ if (ioctl(socketfd, SIOCGIFNAME, &ifr) == -1) {
+ puts("Could not find a non-loopback interface, SKIPPING");
+ return 0;
+ }
+ if (strncmp(ifr.ifr_name, "lo", IFNAMSIZ) != 0) {
+ break;
+ }
+ }
+ printf("Interface index: %d\nInterface name: %.*s\n",
+ ifr.ifr_ifindex, IFNAMSIZ, ifr.ifr_name);
+
+ basic_test(socketfd, ifr);
+
+ n_stats = test_get_sset_info(socketfd, ifr);
+ if (n_stats > 0) {
+ /* Testing lexible arrays. */
+ test_get_stats(socketfd, ifr, n_stats);
+ test_get_strings(socketfd, ifr, n_stats);
+ }
+
+ /* Testing manual implementations of structure convertions. */
+ test_get_rxfh(socketfd, ifr);
+ test_get_link_settings(socketfd, ifr);
+ test_perqueue(socketfd, ifr);
+
+ /* Testing manual implementations of operations. */
+ test_get_rssh(socketfd, ifr);
+
+ return 0;
+}
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 0/6] fcntl, sockopt, and ioctl options
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
` (5 preceding siblings ...)
2020-07-23 0:19 ` [PATCH 6/6] linux-user: Add support for SIOCETHTOOL ioctl Shu-Chun Weng
@ 2020-08-05 23:22 ` Shu-Chun Weng
6 siblings, 0 replies; 11+ messages in thread
From: Shu-Chun Weng @ 2020-08-05 23:22 UTC (permalink / raw)
To: qemu-devel; +Cc: Laurent Vivier
[-- Attachment #1.1: Type: text/plain, Size: 2890 bytes --]
Ping! Patchew: https://patchew.org/QEMU/cover.1595461447.git.scw@google.com/
On Wed, Jul 22, 2020 at 5:19 PM Shu-Chun Weng <scw@google.com> wrote:
> Hi Laurent,
>
> This is a series of 6 patches in 4 groups, putting into a single thread for
> easier tracking.
>
> [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls
> An incidental follow up on
> https://lists.nongnu.org/archive/html/qemu-devel/2019-09/msg01925.html
>
> [PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt
> Updated
> https://lists.nongnu.org/archive/html/qemu-devel/2019-09/msg01317.html
> to consistently add them in get/setsockopt
>
> [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW
> [PATCH 4/6] linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING
> Updated
> https://lists.nongnu.org/archive/html/qemu-devel/2019-09/msg01319.html
> to only use TARGET_SO_*_OLD/NEW
>
> [PATCH 5/6] thunk: supports flexible arrays
> [PATCH 6/6] linux-user: Add support for SIOCETHTOOL ioctl
> Updated
> https://lists.nongnu.org/archive/html/qemu-devel/2019-08/msg05090.html
>
> Shu-Chun Weng (6):
> linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls
> linux-user: add missing UDP and IPv6 get/setsockopt options
> linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW
> linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING
> thunk: supports flexible arrays
> linux-user: Add support for SIOCETHTOOL ioctl
>
> include/exec/user/thunk.h | 20 +
> linux-user/Makefile.objs | 3 +-
> linux-user/alpha/sockbits.h | 21 +-
> linux-user/ethtool.c | 819 +++++++++++++++++++++++++
> linux-user/ethtool.h | 19 +
> linux-user/ethtool_entries.h | 107 ++++
> linux-user/fd-trans.h | 41 +-
> linux-user/generic/sockbits.h | 17 +-
> linux-user/hppa/sockbits.h | 20 +-
> linux-user/ioctls.h | 2 +
> linux-user/mips/sockbits.h | 16 +-
> linux-user/qemu.h | 1 +
> linux-user/sparc/sockbits.h | 21 +-
> linux-user/strace.c | 19 +-
> linux-user/syscall.c | 233 ++++++-
> linux-user/syscall_defs.h | 26 +-
> linux-user/syscall_types.h | 277 +++++++++
> tests/tcg/multiarch/ethtool.c | 417 +++++++++++++
> tests/tcg/multiarch/socket_timestamp.c | 542 ++++++++++++++++
> thunk.c | 151 ++++-
> 20 files changed, 2706 insertions(+), 66 deletions(-)
> create mode 100644 linux-user/ethtool.c
> create mode 100644 linux-user/ethtool.h
> create mode 100644 linux-user/ethtool_entries.h
> create mode 100644 tests/tcg/multiarch/ethtool.c
> create mode 100644 tests/tcg/multiarch/socket_timestamp.c
>
> --
> 2.28.0.rc0.105.gf9edc3c819-goog
>
>
[-- Attachment #1.2: Type: text/html, Size: 4122 bytes --]
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 3844 bytes --]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls
2020-07-23 0:19 ` [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls Shu-Chun Weng
@ 2020-08-07 11:44 ` Laurent Vivier
0 siblings, 0 replies; 11+ messages in thread
From: Laurent Vivier @ 2020-08-07 11:44 UTC (permalink / raw)
To: Shu-Chun Weng, qemu-devel
Le 23/07/2020 à 02:19, Shu-Chun Weng a écrit :
> Signed-off-by: Shu-Chun Weng <scw@google.com>
> ---
> linux-user/syscall.c | 10 ++++++++++
> linux-user/syscall_defs.h | 14 ++++++++------
> 2 files changed, 18 insertions(+), 6 deletions(-)
>
> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
> index 1211e759c2..f97337b0b4 100644
> --- a/linux-user/syscall.c
> +++ b/linux-user/syscall.c
> @@ -6312,6 +6312,14 @@ static int target_to_host_fcntl_cmd(int cmd)
> case TARGET_F_GETPIPE_SZ:
> ret = F_GETPIPE_SZ;
> break;
> +#endif
> +#ifdef F_ADD_SEALS
> + case TARGET_F_ADD_SEALS:
> + ret = F_ADD_SEALS;
> + break;
> + case TARGET_F_GET_SEALS:
> + ret = F_GET_SEALS;
> + break;
> #endif
> default:
> ret = -TARGET_EINVAL;
> @@ -6598,6 +6606,8 @@ static abi_long do_fcntl(int fd, int cmd, abi_ulong arg)
> case TARGET_F_GETLEASE:
> case TARGET_F_SETPIPE_SZ:
> case TARGET_F_GETPIPE_SZ:
> + case TARGET_F_ADD_SEALS:
> + case TARGET_F_GET_SEALS:
> ret = get_errno(safe_fcntl(fd, host_cmd, arg));
> break;
>
> diff --git a/linux-user/syscall_defs.h b/linux-user/syscall_defs.h
> index 3c261cff0e..70df1a94fb 100644
> --- a/linux-user/syscall_defs.h
> +++ b/linux-user/syscall_defs.h
> @@ -2292,12 +2292,14 @@ struct target_statfs64 {
> #endif
>
> #define TARGET_F_LINUX_SPECIFIC_BASE 1024
> -#define TARGET_F_SETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 0)
> -#define TARGET_F_GETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 1)
> -#define TARGET_F_DUPFD_CLOEXEC (TARGET_F_LINUX_SPECIFIC_BASE + 6)
> -#define TARGET_F_SETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 7)
> -#define TARGET_F_GETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 8)
> -#define TARGET_F_NOTIFY (TARGET_F_LINUX_SPECIFIC_BASE+2)
> +#define TARGET_F_SETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 0)
> +#define TARGET_F_GETLEASE (TARGET_F_LINUX_SPECIFIC_BASE + 1)
> +#define TARGET_F_DUPFD_CLOEXEC (TARGET_F_LINUX_SPECIFIC_BASE + 6)
> +#define TARGET_F_NOTIFY (TARGET_F_LINUX_SPECIFIC_BASE + 2)
> +#define TARGET_F_SETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 7)
> +#define TARGET_F_GETPIPE_SZ (TARGET_F_LINUX_SPECIFIC_BASE + 8)
> +#define TARGET_F_ADD_SEALS (TARGET_F_LINUX_SPECIFIC_BASE + 9)
> +#define TARGET_F_GET_SEALS (TARGET_F_LINUX_SPECIFIC_BASE + 10)
>
> #include "target_fcntl.h"
>
>
Looks good. Please also update print_fcntl() in linux-user/strace.c
Thanks,
Laurent
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt options
2020-07-23 0:19 ` [PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt options Shu-Chun Weng
@ 2020-08-07 11:54 ` Laurent Vivier
0 siblings, 0 replies; 11+ messages in thread
From: Laurent Vivier @ 2020-08-07 11:54 UTC (permalink / raw)
To: Shu-Chun Weng; +Cc: qemu-devel
Le 23/07/2020 à 02:19, Shu-Chun Weng a écrit :
> UDP: SOL_UDP manipulate options at UDP level. All six options currently
> defined in linux source include/uapi/linux/udp.h take integer values.
>
> IPv6: IPV6_ADDR_PREFERENCES (RFC5014: Source address selection) was not
> supported.
>
> Signed-off-by: Shu-Chun Weng <scw@google.com>
> ---
> linux-user/syscall.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
This might be really clearer if you split this patch in two: one to add
SOL_UDP and one to add IPV6_ADDR_PREFERENCES.
Also update do_print_sockopt() in linux-use/strace.c
With that done, you can add my "Reviewed-by: Laurent Vivier
<laurent@vivier.eu>" to both.
Thanks,
Laurent
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW
2020-07-23 0:19 ` [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW Shu-Chun Weng
@ 2020-08-07 12:07 ` Laurent Vivier
0 siblings, 0 replies; 11+ messages in thread
From: Laurent Vivier @ 2020-08-07 12:07 UTC (permalink / raw)
To: Shu-Chun Weng, QEMU Developers
Le 23/07/2020 à 02:19, Shu-Chun Weng a écrit :
> Both guest options map to host SO_TIMESTAMP while keeping a bit in
> fd_trans to remember if the guest expects the old or the new format.
I don't think we need to keep this information for each fd.
Once a program has used the _NEW version it will always use the _NEW
version. It's possible to mix, but I don't think we have to support
this. This adds too much complexity.
> Added a multiarch test to verify.
>
> Signed-off-by: Shu-Chun Weng <scw@google.com>
> ---
> linux-user/alpha/sockbits.h | 8 +-
> linux-user/fd-trans.h | 41 +++-
> linux-user/generic/sockbits.h | 9 +-
> linux-user/hppa/sockbits.h | 8 +-
> linux-user/mips/sockbits.h | 8 +-
> linux-user/sparc/sockbits.h | 8 +-
> linux-user/strace.c | 7 +-(fd)
> linux-user/syscall.c | 69 ++++--
> tests/tcg/multiarch/socket_timestamp.c | 292 +++++++++++++++++++++++++
> 9 files changed, 419 insertions(+), 31 deletions(-)
> create mode 100644 tests/tcg/multiarch/socket_timestamp.c
>
Thanks,
Laurent
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2020-08-07 12:08 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-07-23 0:19 [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 1/6] linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls Shu-Chun Weng
2020-08-07 11:44 ` Laurent Vivier
2020-07-23 0:19 ` [PATCH 2/6] linux-user: add missing UDP and IPv6 get/setsockopt options Shu-Chun Weng
2020-08-07 11:54 ` Laurent Vivier
2020-07-23 0:19 ` [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW Shu-Chun Weng
2020-08-07 12:07 ` Laurent Vivier
2020-07-23 0:19 ` [PATCH 4/6] linux-user: setsockopt() SO_TIMESTAMPNS and SO_TIMESTAMPING Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 5/6] thunk: supports flexible arrays Shu-Chun Weng
2020-07-23 0:19 ` [PATCH 6/6] linux-user: Add support for SIOCETHTOOL ioctl Shu-Chun Weng
2020-08-05 23:22 ` [PATCH 0/6] fcntl, sockopt, and ioctl options Shu-Chun Weng
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.