[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] Migration bug added by commit 2df1aa01bef7366798248ac6d03cfb42048b003d



commit 2df1aa01bef7366798248ac6d03cfb42048b003d
Author: Paul Durrant <paul.durrant@xxxxxxxxxx>
Date:   Tue Jun 23 18:07:49 2015 +0200

    x86/hvm: remove hvm_io_pending() check in hvmemul_do_io()

...
-            rc = X86EMUL_RETRY;
-            if ( !hvm_send_assist_req(s, &p) )
+            rc = hvm_send_assist_req(s, &p);
+            if ( rc != X86EMUL_RETRY )
...
     if ( unlikely(!vcpu_start_shutdown_deferral(curr)) )
-        return 0; /* implicitly bins the i/o operation */
+        return X86EMUL_OKAY;

So now X86EMUL_OKAY is returned from hvmemul_do_io() during shutdown.

From Jan Beulich about this:


 Re: [Xen-devel] [PATCH 3/5] hvmemul_do_io: If the send to the ioreq
 server failed do not retry.
 
<https://www.mail-archive.com/search?l=xen-devel@xxxxxxxxxxxxx&q=subject:%22Re%5C%3A+%5C%5BXen%5C-devel%5C%5D+%5C%5BPATCH+3%5C%2F5%5C%5D+hvmemul_do_io%5C%3A+If+the+send+to+the+ioreq+server+failed+do+not+retry.%22&o=newest>


Jan Beulich <https://www.mail-archive.com/search?l=xen-devel@xxxxxxxxxxxxx&q=from:%22Jan+Beulich%22> Fri, 30 Jan 2015 02:24:55 -0800 <https://www.mail-archive.com/search?l=xen-devel@xxxxxxxxxxxxx&q=date:20150130>

On 30.01.15 at 01:52, <dsl...@xxxxxxxxxxx> wrote:
I.E. do just what no backing DM does.


_If_ this is correct, the if() modified here should be folded with the
one a few lines up. But looking at the description of the commit that
introduced this (bac0999325 "x86 hvm: Do not incorrectly retire an
instruction emulation...", almost immediately modified by f20f3c8ece
"x86 hvm: On failed hvm_send_assist_req(), io emulation...") I doubt
this is really what we want, or at the very least your change
description should explain what was wrong with the original commit.

Jan



going up the call stack:

in handle_pio when hvmemul_do_pio_buffer() returns X86EMUL_OKAY, it returns 1.

   svm_vmexit_handler 2578 if ( handle_pio(port, bytes, dir) )
or
   vmx_vmexit_handler 3178 if ( handle_pio(port, bytes, dir) )

both update the IP in this case which is wrong during shutdown.

So I think:

            rc = hvm_send_assist_req(s, &p);
            if ( rc != X86EMUL_RETRY )
                 vio->io_state = HVMIO_none;

needs to change to:

                rc = hvm_send_assist_req(s, &p);
                if ( rc != X86EMUL_RETRY )
                {
                    vio->io_state = HVMIO_none;
                    if ( rc == X86EMUL_OKAY )
                        rc = X86EMUL_RETRY;
                }


   -Don Slutz

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.