All of lore.kernel.org
 help / color / mirror / Atom feed
* [uml-devel] VDE backend does not handle EINTR properly
@ 2011-11-19 23:16 Frank Laub
  2011-11-19 23:27 ` Richard Weinberger
  0 siblings, 1 reply; 3+ messages in thread
From: Frank Laub @ 2011-11-19 23:16 UTC (permalink / raw)
  To: user-mode-linux-devel

When running the VDE backend with 2 uml instances and running iperf 
between them, the vde_send() function often returns EINTR. I noticed 
that this is accounted for in net_user.c via the use of the 
CATCH_EINTR() macro. Thus the following patch uses the same approach for 
calls into VDE. It could be argued that the VDE library itself should 
handle EINTR, but this patch will work with older (or current) versions 
of VDE. This patch was created against 3.0.3.

--- a/arch/um/drivers/vde_user.c
+++ b/arch/um/drivers/vde_user.c
@@ -11,6 +11,7 @@
  #include "um_malloc.h"
  #include "user.h"
  #include "vde.h"
+#include "os.h"

  static int vde_user_init(void *data, void *dev)
  {
@@ -103,7 +104,7 @@ int vde_user_read(void *conn, void *buf,
      if (vconn == NULL)
          return 0;

-    rv = vde_recv(vconn, buf, len, 0);
+    CATCH_EINTR(rv = vde_recv(vconn, buf, len, 0));
      if (rv < 0) {
          if (errno == EAGAIN)
              return 0;
@@ -118,10 +119,19 @@ int vde_user_read(void *conn, void *buf,
  int vde_user_write(void *conn, void *buf, int len)
  {
      VDECONN *vconn = conn;
+    int rv;

      if (vconn == NULL)
          return 0;

-    return vde_send(vconn, buf, len, 0);
+    CATCH_EINTR(rv = vde_send(vconn, buf, len, 0));
+    if (rv < 0) {
+        if (errno == EAGAIN)
+            return 0;
+        return -errno;
+    }
+    else if (rv == 0)
+        return -ENOTCONN;
+    return rv;
  }



------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [uml-devel] VDE backend does not handle EINTR properly
  2011-11-19 23:16 [uml-devel] VDE backend does not handle EINTR properly Frank Laub
@ 2011-11-19 23:27 ` Richard Weinberger
  2011-11-20  1:29   ` [uml-devel] ]PATCH 3.03 1/1] um: VDE backend needs to handle EINTR/EAGAIN Frank Laub
  0 siblings, 1 reply; 3+ messages in thread
From: Richard Weinberger @ 2011-11-19 23:27 UTC (permalink / raw)
  To: Frank Laub; +Cc: user-mode-linux-devel

Am 20.11.2011 00:16, schrieb Frank Laub:
> When running the VDE backend with 2 uml instances and running iperf 
> between them, the vde_send() function often returns EINTR. I noticed 
> that this is accounted for in net_user.c via the use of the 
> CATCH_EINTR() macro. Thus the following patch uses the same approach for 
> calls into VDE. It could be argued that the VDE library itself should 
> handle EINTR, but this patch will work with older (or current) versions 
> of VDE. This patch was created against 3.0.3.

Please add a Signed-off-by tag,
so that I can give you the credit. :-)
See: http://linux.yyz.us/patch-format.html

> --- a/arch/um/drivers/vde_user.c
> +++ b/arch/um/drivers/vde_user.c
> @@ -11,6 +11,7 @@
>   #include "um_malloc.h"
>   #include "user.h"
>   #include "vde.h"
> +#include "os.h"
> 
>   static int vde_user_init(void *data, void *dev)
>   {
> @@ -103,7 +104,7 @@ int vde_user_read(void *conn, void *buf,
>       if (vconn == NULL)
>           return 0;
> 
> -    rv = vde_recv(vconn, buf, len, 0);
> +    CATCH_EINTR(rv = vde_recv(vconn, buf, len, 0));
>       if (rv < 0) {
>           if (errno == EAGAIN)
>               return 0;
> @@ -118,10 +119,19 @@ int vde_user_read(void *conn, void *buf,
>   int vde_user_write(void *conn, void *buf, int len)
>   {
>       VDECONN *vconn = conn;
> +    int rv;
> 
>       if (vconn == NULL)
>           return 0;
> 
> -    return vde_send(vconn, buf, len, 0);

Is the additional if-branch really needed?

> +    CATCH_EINTR(rv = vde_send(vconn, buf, len, 0));
> +    if (rv < 0) {
> +        if (errno == EAGAIN)
> +            return 0;
> +        return -errno;
> +    }
> +    else if (rv == 0)
> +        return -ENOTCONN;
> +    return rv;
>   }
> 

Thanks,
//richard

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [uml-devel] ]PATCH 3.03 1/1] um: VDE backend needs to handle EINTR/EAGAIN
  2011-11-19 23:27 ` Richard Weinberger
@ 2011-11-20  1:29   ` Frank Laub
  0 siblings, 0 replies; 3+ messages in thread
From: Frank Laub @ 2011-11-20  1:29 UTC (permalink / raw)
  Cc: user-mode-linux-devel

Handle EINTR and EAGAIN during vde_send() and vde_recv() in case the VDE library doesn't. Use the same approach that net_user.c takes.

Signed-off-by: Frank Laub<flaub@anomali.es>

--- a/arch/um/drivers/vde_user.c
+++ b/arch/um/drivers/vde_user.c
@@ -11,6 +11,7 @@
  #include "um_malloc.h"
  #include "user.h"
  #include "vde.h"
+#include "os.h"

  static int vde_user_init(void *data, void *dev)
  {
@@ -103,7 +104,7 @@ int vde_user_read(void *conn, void *buf,
      if (vconn == NULL)
          return 0;

-    rv = vde_recv(vconn, buf, len, 0);
+    CATCH_EINTR(rv = vde_recv(vconn, buf, len, 0));
      if (rv < 0) {
          if (errno == EAGAIN)
              return 0;
@@ -118,10 +119,17 @@ int vde_user_read(void *conn, void *buf,
  int vde_user_write(void *conn, void *buf, int len)
  {
      VDECONN *vconn = conn;
+    int rv;

      if (vconn == NULL)
          return 0;

-    return vde_send(vconn, buf, len, 0);
+    CATCH_EINTR(rv = vde_send(vconn, buf, len, 0));
+    if (rv < 0) {
+        if (errno == EAGAIN)
+            return 0;
+        return -errno;
+    }
+    return rv;
  }


------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2011-11-20  1:29 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-11-19 23:16 [uml-devel] VDE backend does not handle EINTR properly Frank Laub
2011-11-19 23:27 ` Richard Weinberger
2011-11-20  1:29   ` [uml-devel] ]PATCH 3.03 1/1] um: VDE backend needs to handle EINTR/EAGAIN Frank Laub

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.