Bug 231444

Summary: dhclient broken in Xen guest - need xen-checksum patch
Product: [Fedora] Fedora Reporter: Mark McLoughlin <markmc>
Component: dhcpAssignee: David Cantrell <dcantrell>
Status: CLOSED RAWHIDE QA Contact:
Severity: medium Docs Contact:
Priority: medium    
Version: rawhideCC: bstein, dcantrell, herbert.xu, xen-maint
Target Milestone: ---   
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2007-04-11 10:30:13 EDT Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Attachments:
Description Flags
linux-2.6.19-packet-auxdata-csum.patch
none
dhcp-3.0.5-xen-checksum.patch
none
dhcp-3.0.5-xen-checksum.patch none

Description Mark McLoughlin 2007-03-08 07:25:46 EST
kernel-xen-2.6.19-1.2898.2.3.fc7
dhclient-3.0.5-25.2.fc7

Basically, running dhclient in a Xen guest, dhclient is ignoring DHCP relies and
I'm seeing this in syslog:

Mar  7 12:11:29 localhost dhclient: 5 bad udp checksums in 5 packets

The background to this is that packets from Xen guests on the same host only
have partial checksums. (Not sure why)

In rh #219681, Herbert proposed a set of patches to fix this - first make the
TP_STATUS_CSUMNOTREADY flag available via recvmsg() auxdata and, second, make
dhclient check this flag and ignore the checksums if its set.

These patches aren't in rawhide, but I've tried applying the patches and it
seems dhclient isn't seeing the replies at all now - i.e. I'm seeing this in the
strace:

select(7, [5 6], [], [], {7, 996967})   = 0 (Timeout)
gettimeofday({1173356524, 521}, NULL)   = 0
time([1173356524])                      = 1173356524
stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=3533, ...}) = 0
stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=3533, ...}) = 0
stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=3533, ...}) = 0
sendto(3, "<30>Mar  8 12:22:04 dhclient: DH"..., 89, MSG_NOSIGNAL, NULL, 0) = 89
write(2, "DHCPDISCOVER on eth0 to 255.255."..., 59) = 59
write(2, "\n", 1)                       = 1
write(6, "\377\377\377\377\377\377\0\26>l\5\0\10\0E\20\1H\0\0\0\0"..., 342) = 342
gettimeofday({1173356524, 1094}, NULL)  = 0
select(7, [5 6], [], [], {11, 998906})  = 0 (Timeout)



Attaching the patches as I've built them.
Comment 1 Mark McLoughlin 2007-03-08 07:28:04 EST
Created attachment 149561 [details]
linux-2.6.19-packet-auxdata-csum.patch
Comment 2 Mark McLoughlin 2007-03-08 07:28:58 EST
Created attachment 149562 [details]
dhcp-3.0.5-xen-checksum.patch
Comment 3 Mark McLoughlin 2007-03-08 07:35:30 EST
(Sorry, bug #219681 is the link to the RHEL5 bug)
Comment 4 Mark McLoughlin 2007-04-11 10:07:09 EDT
Okay, linux-2.6-xen.patch in rawhide kernel-xen-2.6.20-2925.5.fc7 has Herbert's
original patch and his subsequent fix, both of which were sent upstream here:

  http://www.mail-archive.com/netdev@vger.kernel.org/msg29299.html

So, what's needed is the xen-checksum patch which was removed from rawhide.
Moving to dhcp.
Comment 5 David Cantrell 2007-04-11 10:16:04 EDT
Created attachment 152271 [details]
dhcp-3.0.5-xen-checksum.patch
Comment 6 Mark McLoughlin 2007-04-11 10:29:12 EDT
Both dhcp-3.0.5-xen-checksum.patch patches are identical, both fix the problem
with rawhide kernel and dhclient
Comment 7 David Cantrell 2007-04-11 10:30:13 EDT
Patch applied to dhcp-3.0.5-30.fc7 in rawhide.  Thanks.