Re: [Xen-devel] [PATCH RFC V2 7/7] COLO-Proxy: Use socket to get checkpoint event.

2017-02-15 Thread Zhang Chen 



On 02/16/2017 12:24 AM, Konrad Rzeszutek Wilk wrote:

On Wed, Feb 15, 2017 at 05:54:33PM +0800, Zhang Chen wrote:

We use kernel colo proxy's way to get the checkpoint event
from qemu colo-compare.
Qemu colo-compare need add a API to support this(I will add this in qemu).

Signed-off-by: Zhang Chen 
---
  tools/libxl/libxl_colo.h |  2 +
  tools/libxl/libxl_colo_proxy.c   | 84 +---
  tools/libxl/libxl_colo_restore.c | 11 --
  tools/libxl/libxl_colo_save.c| 22 +++
  tools/libxl/libxl_nic.c  |  4 ++
  tools/libxl/libxl_types.idl  |  4 +-
  tools/libxl/xl_cmdimpl.c |  4 ++
  7 files changed, 113 insertions(+), 18 deletions(-)

diff --git a/tools/libxl/libxl_colo.h b/tools/libxl/libxl_colo.h
index 4746d8c..6c01b55 100644
--- a/tools/libxl/libxl_colo.h
+++ b/tools/libxl/libxl_colo.h
@@ -69,6 +69,8 @@ struct libxl__colo_proxy_state {
   *  False means use kernel colo proxy.
   */
  bool is_userspace_proxy;
+const char *checkpoint_host;
+const char *checkpoint_port;
  };
  
  struct libxl__colo_save_state {

diff --git a/tools/libxl/libxl_colo_proxy.c b/tools/libxl/libxl_colo_proxy.c
index dd902fc..9d21cf1 100644
--- a/tools/libxl/libxl_colo_proxy.c
+++ b/tools/libxl/libxl_colo_proxy.c
@@ -18,6 +18,9 @@
  #include "libxl_internal.h"
  
  #include 

+#include 
+#include 
+#include 
  
  /* Consistent with the new COLO netlink channel in kernel side */

  #define NETLINK_COLO 28
@@ -76,6 +79,26 @@ static int colo_proxy_send(libxl__colo_proxy_state *cps, 
uint8_t *buff,
  return ret;
  }
  
+static int colo_userspace_proxy_recv(libxl__colo_proxy_state *cps,

+ char *buff,
+ unsigned int timeout_us)
+{
+struct timeval tv;
+int ret;
+
+STATE_AO_GC(cps->ao);
+
+if (timeout_us) {
+tv.tv_sec = timeout_us / 100;
+tv.tv_usec = timeout_us % 100;
+setsockopt(cps->sock_fd, SOL_SOCKET, SO_RCVTIMEO, , sizeof(tv));
+}
+
+ret = recv(cps->sock_fd, buff, sizeof(buff),0);
+
+return ret;
+}
+
  /* error: return -1, otherwise return 0 */
  static int64_t colo_proxy_recv(libxl__colo_proxy_state *cps, uint8_t **buff,
 unsigned int timeout_us)
@@ -153,8 +176,45 @@ int colo_proxy_setup(libxl__colo_proxy_state *cps)
  STATE_AO_GC(cps->ao);
  
  /* If enable userspace proxy mode, we don't need setup kernel proxy */

-if (cps->is_userspace_proxy)
+if (cps->is_userspace_proxy) {
+struct sockaddr_in addr;
+int port;
+char recvbuff[1024];
+
+memset(, 0, sizeof(addr));
+port = atoi(cps->checkpoint_port);
+addr.sin_family = AF_INET;
+addr.sin_port = htons(port);
+addr.sin_addr.s_addr = inet_addr(cps->checkpoint_host);
+
+skfd = socket(AF_INET, SOCK_STREAM, 0);
+if (skfd < 0) {
+LOGD(ERROR, ao->domid, "can not create a TCP socket: %s",
+ strerror(errno));
+goto out;
+}
+
+cps->sock_fd = skfd;
+
+if (connect(skfd, (struct sockaddr *), sizeof(addr)) < 0) {
+LOGD(ERROR, ao->domid, "connect error");
+goto out;
+}
+
+char sendbuf[] = "COLO_USERSPACE_PROXY_INIT";

Um, could you move it right around 'char recvbuff' ?


OK, I will move it in next version.


+ret = send(skfd, sendbuf, strlen(sendbuf),0);
+if (ret < 0)
+goto out;
+
+ret = colo_userspace_proxy_recv(cps, recvbuff, 50);

That 50 looks like a good candidate for a #define?


I will add a macro "#define COLO_DEFAULT_WAIT_TIME 50".

Thanks
Zhang Chen




+if (size < 0) {


s/size/ret


+LOGD(ERROR, ao->domid, "Can't recv msg from qemu colo-compare: %s",
+ strerror(errno));
+goto out;
+}
+
  return 0;
+}
  
  skfd = socket(PF_NETLINK, SOCK_RAW, NETLINK_COLO);

  if (skfd < 0) {
@@ -247,8 +307,11 @@ void colo_proxy_preresume(libxl__colo_proxy_state *cps)
   * If enable userspace proxy mode,
   * we don't need preresume kernel proxy
   */
-if (cps->is_userspace_proxy)
+if (cps->is_userspace_proxy) {
+char sendbuf[] = "COLO_CHECKPOINT";
+send(cps->sock_fd, sendbuf, strlen(sendbuf),0);
  return;
+}
  
  colo_proxy_send(cps, NULL, 0, COLO_CHECKPOINT);

  /* TODO: need to handle if the call fails... */
@@ -277,16 +340,25 @@ int colo_proxy_checkpoint(libxl__colo_proxy_state *cps,
  struct nlmsghdr *h;
  struct colo_msg *m;
  int ret = -1;
+char recvbuff[1024];
  
  STATE_AO_GC(cps->ao);
  
  /*

- * enable userspace proxy mode, tmp sleep.
- * then we will add qemu API support this func.
+ * enable userspace proxy mode.
+ * Then we will add qemu API support for this func.
   */
  if

Re: [Xen-devel] [PATCH RFC V2 7/7] COLO-Proxy: Use socket to get checkpoint event.

2017-02-15 Thread Konrad Rzeszutek Wilk 
On Wed, Feb 15, 2017 at 05:54:33PM +0800, Zhang Chen wrote:
> We use kernel colo proxy's way to get the checkpoint event
> from qemu colo-compare.
> Qemu colo-compare need add a API to support this(I will add this in qemu).
> 
> Signed-off-by: Zhang Chen 
> ---
>  tools/libxl/libxl_colo.h |  2 +
>  tools/libxl/libxl_colo_proxy.c   | 84 
> +---
>  tools/libxl/libxl_colo_restore.c | 11 --
>  tools/libxl/libxl_colo_save.c| 22 +++
>  tools/libxl/libxl_nic.c  |  4 ++
>  tools/libxl/libxl_types.idl  |  4 +-
>  tools/libxl/xl_cmdimpl.c |  4 ++
>  7 files changed, 113 insertions(+), 18 deletions(-)
> 
> diff --git a/tools/libxl/libxl_colo.h b/tools/libxl/libxl_colo.h
> index 4746d8c..6c01b55 100644
> --- a/tools/libxl/libxl_colo.h
> +++ b/tools/libxl/libxl_colo.h
> @@ -69,6 +69,8 @@ struct libxl__colo_proxy_state {
>   *  False means use kernel colo proxy.
>   */
>  bool is_userspace_proxy;
> +const char *checkpoint_host;
> +const char *checkpoint_port;
>  };
>  
>  struct libxl__colo_save_state {
> diff --git a/tools/libxl/libxl_colo_proxy.c b/tools/libxl/libxl_colo_proxy.c
> index dd902fc..9d21cf1 100644
> --- a/tools/libxl/libxl_colo_proxy.c
> +++ b/tools/libxl/libxl_colo_proxy.c
> @@ -18,6 +18,9 @@
>  #include "libxl_internal.h"
>  
>  #include 
> +#include 
> +#include 
> +#include 
>  
>  /* Consistent with the new COLO netlink channel in kernel side */
>  #define NETLINK_COLO 28
> @@ -76,6 +79,26 @@ static int colo_proxy_send(libxl__colo_proxy_state *cps, 
> uint8_t *buff,
>  return ret;
>  }
>  
> +static int colo_userspace_proxy_recv(libxl__colo_proxy_state *cps,
> + char *buff,
> + unsigned int timeout_us)
> +{
> +struct timeval tv;
> +int ret;
> +
> +STATE_AO_GC(cps->ao);
> +
> +if (timeout_us) {
> +tv.tv_sec = timeout_us / 100;
> +tv.tv_usec = timeout_us % 100;
> +setsockopt(cps->sock_fd, SOL_SOCKET, SO_RCVTIMEO, , sizeof(tv));
> +}
> +
> +ret = recv(cps->sock_fd, buff, sizeof(buff),0);
> +
> +return ret;
> +}
> +
>  /* error: return -1, otherwise return 0 */
>  static int64_t colo_proxy_recv(libxl__colo_proxy_state *cps, uint8_t **buff,
> unsigned int timeout_us)
> @@ -153,8 +176,45 @@ int colo_proxy_setup(libxl__colo_proxy_state *cps)
>  STATE_AO_GC(cps->ao);
>  
>  /* If enable userspace proxy mode, we don't need setup kernel proxy */
> -if (cps->is_userspace_proxy)
> +if (cps->is_userspace_proxy) {
> +struct sockaddr_in addr;
> +int port;
> +char recvbuff[1024];
> +
> +memset(, 0, sizeof(addr));
> +port = atoi(cps->checkpoint_port);
> +addr.sin_family = AF_INET;
> +addr.sin_port = htons(port);
> +addr.sin_addr.s_addr = inet_addr(cps->checkpoint_host);
> +
> +skfd = socket(AF_INET, SOCK_STREAM, 0);
> +if (skfd < 0) {
> +LOGD(ERROR, ao->domid, "can not create a TCP socket: %s",
> + strerror(errno));
> +goto out;
> +}
> +
> +cps->sock_fd = skfd;
> +
> +if (connect(skfd, (struct sockaddr *), sizeof(addr)) < 0) {
> +LOGD(ERROR, ao->domid, "connect error");
> +goto out;
> +}
> +
> +char sendbuf[] = "COLO_USERSPACE_PROXY_INIT";

Um, could you move it right around 'char recvbuff' ?

> +ret = send(skfd, sendbuf, strlen(sendbuf),0);
> +if (ret < 0)
> +goto out;
> +
> +ret = colo_userspace_proxy_recv(cps, recvbuff, 50);

That 50 looks like a good candidate for a #define?

> +if (size < 0) {
> +LOGD(ERROR, ao->domid, "Can't recv msg from qemu colo-compare: 
> %s",
> + strerror(errno));
> +goto out;
> +}
> +
>  return 0;
> +}
>  
>  skfd = socket(PF_NETLINK, SOCK_RAW, NETLINK_COLO);
>  if (skfd < 0) {
> @@ -247,8 +307,11 @@ void colo_proxy_preresume(libxl__colo_proxy_state *cps)
>   * If enable userspace proxy mode,
>   * we don't need preresume kernel proxy
>   */
> -if (cps->is_userspace_proxy)
> +if (cps->is_userspace_proxy) {
> +char sendbuf[] = "COLO_CHECKPOINT";
> +send(cps->sock_fd, sendbuf, strlen(sendbuf),0);
>  return;
> +}
>  
>  colo_proxy_send(cps, NULL, 0, COLO_CHECKPOINT);
>  /* TODO: need to handle if the call fails... */
> @@ -277,16 +340,25 @@ int colo_proxy_checkpoint(libxl__colo_proxy_state *cps,
>  struct nlmsghdr *h;
>  struct colo_msg *m;
>  int ret = -1;
> +char recvbuff[1024];
>  
>  STATE_AO_GC(cps->ao);
>  
>  /*
> - * enable userspace proxy mode, tmp sleep.
> - * then we will add qemu API support this func.
> + * enable userspace proxy mode.
> + * Then we will add qemu API

[Xen-devel] [PATCH RFC V2 7/7] COLO-Proxy: Use socket to get checkpoint event.

2017-02-15 Thread Zhang Chen 
We use kernel colo proxy's way to get the checkpoint event
from qemu colo-compare.
Qemu colo-compare need add a API to support this(I will add this in qemu).

Signed-off-by: Zhang Chen 
---
 tools/libxl/libxl_colo.h |  2 +
 tools/libxl/libxl_colo_proxy.c   | 84 +---
 tools/libxl/libxl_colo_restore.c | 11 --
 tools/libxl/libxl_colo_save.c| 22 +++
 tools/libxl/libxl_nic.c  |  4 ++
 tools/libxl/libxl_types.idl  |  4 +-
 tools/libxl/xl_cmdimpl.c |  4 ++
 7 files changed, 113 insertions(+), 18 deletions(-)

diff --git a/tools/libxl/libxl_colo.h b/tools/libxl/libxl_colo.h
index 4746d8c..6c01b55 100644
--- a/tools/libxl/libxl_colo.h
+++ b/tools/libxl/libxl_colo.h
@@ -69,6 +69,8 @@ struct libxl__colo_proxy_state {
  *  False means use kernel colo proxy.
  */
 bool is_userspace_proxy;
+const char *checkpoint_host;
+const char *checkpoint_port;
 };
 
 struct libxl__colo_save_state {
diff --git a/tools/libxl/libxl_colo_proxy.c b/tools/libxl/libxl_colo_proxy.c
index dd902fc..9d21cf1 100644
--- a/tools/libxl/libxl_colo_proxy.c
+++ b/tools/libxl/libxl_colo_proxy.c
@@ -18,6 +18,9 @@
 #include "libxl_internal.h"
 
 #include 
+#include 
+#include 
+#include 
 
 /* Consistent with the new COLO netlink channel in kernel side */
 #define NETLINK_COLO 28
@@ -76,6 +79,26 @@ static int colo_proxy_send(libxl__colo_proxy_state *cps, 
uint8_t *buff,
 return ret;
 }
 
+static int colo_userspace_proxy_recv(libxl__colo_proxy_state *cps,
+ char *buff,
+ unsigned int timeout_us)
+{
+struct timeval tv;
+int ret;
+
+STATE_AO_GC(cps->ao);
+
+if (timeout_us) {
+tv.tv_sec = timeout_us / 100;
+tv.tv_usec = timeout_us % 100;
+setsockopt(cps->sock_fd, SOL_SOCKET, SO_RCVTIMEO, , sizeof(tv));
+}
+
+ret = recv(cps->sock_fd, buff, sizeof(buff),0);
+
+return ret;
+}
+
 /* error: return -1, otherwise return 0 */
 static int64_t colo_proxy_recv(libxl__colo_proxy_state *cps, uint8_t **buff,
unsigned int timeout_us)
@@ -153,8 +176,45 @@ int colo_proxy_setup(libxl__colo_proxy_state *cps)
 STATE_AO_GC(cps->ao);
 
 /* If enable userspace proxy mode, we don't need setup kernel proxy */
-if (cps->is_userspace_proxy)
+if (cps->is_userspace_proxy) {
+struct sockaddr_in addr;
+int port;
+char recvbuff[1024];
+
+memset(, 0, sizeof(addr));
+port = atoi(cps->checkpoint_port);
+addr.sin_family = AF_INET;
+addr.sin_port = htons(port);
+addr.sin_addr.s_addr = inet_addr(cps->checkpoint_host);
+
+skfd = socket(AF_INET, SOCK_STREAM, 0);
+if (skfd < 0) {
+LOGD(ERROR, ao->domid, "can not create a TCP socket: %s",
+ strerror(errno));
+goto out;
+}
+
+cps->sock_fd = skfd;
+
+if (connect(skfd, (struct sockaddr *), sizeof(addr)) < 0) {
+LOGD(ERROR, ao->domid, "connect error");
+goto out;
+}
+
+char sendbuf[] = "COLO_USERSPACE_PROXY_INIT";
+ret = send(skfd, sendbuf, strlen(sendbuf),0);
+if (ret < 0)
+goto out;
+
+ret = colo_userspace_proxy_recv(cps, recvbuff, 50);
+if (size < 0) {
+LOGD(ERROR, ao->domid, "Can't recv msg from qemu colo-compare: %s",
+ strerror(errno));
+goto out;
+}
+
 return 0;
+}
 
 skfd = socket(PF_NETLINK, SOCK_RAW, NETLINK_COLO);
 if (skfd < 0) {
@@ -247,8 +307,11 @@ void colo_proxy_preresume(libxl__colo_proxy_state *cps)
  * If enable userspace proxy mode,
  * we don't need preresume kernel proxy
  */
-if (cps->is_userspace_proxy)
+if (cps->is_userspace_proxy) {
+char sendbuf[] = "COLO_CHECKPOINT";
+send(cps->sock_fd, sendbuf, strlen(sendbuf),0);
 return;
+}
 
 colo_proxy_send(cps, NULL, 0, COLO_CHECKPOINT);
 /* TODO: need to handle if the call fails... */
@@ -277,16 +340,25 @@ int colo_proxy_checkpoint(libxl__colo_proxy_state *cps,
 struct nlmsghdr *h;
 struct colo_msg *m;
 int ret = -1;
+char recvbuff[1024];
 
 STATE_AO_GC(cps->ao);
 
 /*
- * enable userspace proxy mode, tmp sleep.
- * then we will add qemu API support this func.
+ * enable userspace proxy mode.
+ * Then we will add qemu API support for this func.
  */
 if (cps->is_userspace_proxy) {
-sleep(timeout_us / 100);
-return 0;
+ret = colo_userspace_proxy_recv(cps, recvbuff, timeout_us);
+if (ret <= 0)
+return 0;
+
+if (!strcmp(recvbuff, "DO_CHECKPOINT")) {
+return 1;
+} else {
+LOGD(ERROR, ao->domid, "receive qemu colo-compare checkpoint 
error");
+return -1;
+}
 }