[syslinux] PXE boot hangs while transferring vmlinuz or initrd.img

[Michael Madore]

On Tue, Mar 17, 2009 at 9:35 PM, Bishop <bishopolis at gmail.com> wrote:
> Michael,
>
> Are you using 64-bit or 32- ?  We see the same thing on C5 or RHEL5, HP HW,
> 64 bit, over a large number.  32 doesn't seem to run into this issue.  Is
> yours intermittent?

Hi,

The tftp server is running CentOS 5.2 64-bit.  The user is trying to
PXE boot CentOS 5.2 64-bit.  The boot process gets stuck transferring
either the kernel or initial ramdisk at least half the time.

Are you saying you don't see the problem with a 32-bit tftp server, or
booting a 32-bit client?

Mike

>> Hi,
>>
>> I have a user that is experiencing problems pxe booting several
>> identical servers.  pxelinux.0 is from syslinux 3.63 and the TFTP
>> server is 0.42 running on CentOS 5.2.
>>
>> After loading pxelinux.0, the client will then start transferring
>> vmlinuz and initrd.img via tftp.  More than half the time, the
>> transfer will get stuck and then eventually print a timeout message.
>> I ran tcpdump and I can see that the server stops sending data
>> packets.  The client keeps sending the last ACK until it times out:
>>
>> No.     Time        Source                Destination           Protocol
>> Info
>>   2456 62.303446   10.4.253.5            10.4.253.206          TFTP
>>  Data Packet, Block: 1185
>>   2457 62.303768   10.4.253.206          10.4.253.5            TFTP
>>  Acknowledgement, Block: 1185
>>   2458 62.303782   10.4.253.5            10.4.253.206          TFTP
>>  Data Packet, Block: 1186
>>   2459 62.304105   10.4.253.206          10.4.253.5            TFTP
>>  Acknowledgement, Block: 1186
>>   2460 62.304120   10.4.253.5            10.4.253.206          TFTP
>>  Data Packet, Block: 1187
>>   2461 62.307321   10.4.253.206          10.4.253.5            TFTP
>>  Acknowledgement, Block: 1187
>>   2462 65.561489   10.4.253.206          10.4.253.5            TFTP
>>  Acknowledgement, Block: 1187
>>   2463 72.152335   10.4.253.206          10.4.253.5            TFTP
>>  Acknowledgement, Block: 1187
>>
>> It appears to get stuck in a random location in either vmlinuz or
>> initrd.img.  The strange thing is that this only happens when booting
>> the kernel and ramdisk from CentOS 5.2.  Booting the kernel and
>> ramdisk from CentOS 4 works.
>>
>> Any suggestions for debugging this further?