Can you elaborate which cases have duplicate fds?
I've already used closesocket instead of close handle for sockets. I'll send a
separate patch with the stream related changes. E.g. Replace read/write on
socket with send/recv.
Linda
----- Original Message -----
From: Eitan Eliahu <elia...@vmware.com>
To: Ben Pfaff <b...@nicira.com>
Cc: dev@openvswitch.org, Linda Sun <l...@vmware.com>
Sent: Fri, 06 Dec 2013 16:18:51 -0800 (PST)
Subject: Re: [ovs-dev] [PATCH] Use WaitForMultipleObjects for polling on
windows. This works on all kinds of objects, e.g. sockets, files, especially
ioctl calls to the kernel. One additional paramater is passed down in
poll_fd_wait() to help WaitForMultipleObjects. latch is signaled with event, to
be waited/polled by WaitForMultipleObjects() as well.
Ben and Linda,
It seems that the first two comments could be addressed (by removing NULL
events and duplicates handles.
In regards to the third comments, if these handles are Sockets (which I believe
they are) these handles should never be closed by calling CloseHandle(). For
these socket handles closesocket() should be used and therefore the resources
associated with these handles will be freed and WaitForMultipleObjects will
return with an error.
Thanks,
Eitan
----- Original Message -----
From: "Ben Pfaff" <b...@nicira.com>
To: "Linda Sun" <l...@vmware.com>
Cc: dev@openvswitch.org
Sent: Friday, December 6, 2013 3:57:03 PM
Subject: Re: [ovs-dev] [PATCH] Use WaitForMultipleObjects for polling on
windows. This works on all kinds of objects, e.g. sockets, files, especially
ioctl calls to the kernel. One additional paramater is passed down in
poll_fd_wait() to help WaitForMultipleObjects. latch is signaled with event, to
be waited/polled by WaitForMultipleObjects() as well.
On Fri, Dec 06, 2013 at 02:47:29PM -0800, Linda Sun wrote:
> Signed-off-by: Linda Sun <l...@vmware.com>
Hi Linda. Thanks for sending this along so that we can start figuring
out how to work Windows into the poll loop.
Looking at the documentation for WaitForMultipleObjects, I see a few
possible issues.
One is that it looks like we add a lot of null pointers to the wevents
array (most of the poll_fd_wait() calls pass NULL). I don't see
documentation for what WaitForMultipleObjects does with null pointers.
WaitForMultipleObjects also forbids duplicates in its handles array.
OVS calls poll with duplicates sometimes.
WaitForMultipleObjects documentation says: "If one of these handles is
closed while the wait is still pending, the function's behavior is
undefined." OVS might sometimes close a file descriptor from one
thread that some other thread is waiting for, since Unix has
well-defined semantics for that. I'm not sure how to audit the code
for this, or what the consequences for getting it wrong are.
("Undefined" sounds bad, but what happens in practice?)
Thanks,
Ben.
_______________________________________________
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v1/url?u=http://openvswitch.org/mailman/listinfo/dev&k=oIvRg1%2BdGAgOoM1BIlLLqw%3D%3D%0A&r=yTvML8OxA42Jb6ViHe7fUXbvPVOYDPVq87w43doxtlY%3D%0A&m=7pHyuyR919nucNKNtulWpl9q3u58pd6CW4vv29IF5NY%3D%0A&s=3708003efa1c031072b6af10a1c385a2328f4c5f5a6732040e544b86cb0194dd
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
