[ipxe-devel] EFI entropy commit

Jarrod Johnson jarrod.b.johnson at gmail.com
Fri Nov 1 18:24:15 UTC 2013

FWIW this is on IBM hardware, and it has gone 5 times in a row managing to
print '2984' as the last sample out the serial port before rebooting.  I
reduced to 2048 and it said tests passed at the end, though then it seems
to pause and just say '2056 tests passed'  and then the iPXE banner with
features listed.  Was there an expectation of a summary or just the samples?

vmware guest in efi mode gets much much further but also seems to reboot
suddenly (wasn't logging console output).  If I restrict sample count there
it behaves the same as baremetal.

I found that if I disabled the serial console redirection, the system gets
further.  I'm wondering if the serial output was lagging and it is crashing
at the next block generation....

On Fri, Nov 1, 2013 at 1:54 PM, Michael Brown <mcb30 at ipxe.org> wrote:

> On 01/11/13 17:49, Jarrod Johnson wrote:
>> So my console logging catches 2984 samples and the system then reboots
>> abruptly.  I can send a file with 2984 samples explicitly enumerated,
>> but not sure if there was more to it that I was missing.  It suddenly
>> reboots the system without further output after 2984 samples (wel, I
>> guess 0-2984).
> Interesting.  The code in entropy_sample.c should generate 65536 samples
> in blocks of 256 samples.  Rebooting after 2984 samples (i.e. not on a
> 256-sample boundary) suggests that it reboots in the middle of the trivial
> printf() loop:
>         /* Print out sample values */
>         for ( j = 0 ; j < SAMPLE_BLOCKSIZE ; j++ ) {
>                 printf ( "SAMPLE %d %d\n", ( i * SAMPLE_BLOCKSIZE + j ),
>                          samples[j] );
>         }
> I can't see any reason why this should cause a reboot.  This sounds like a
> bug somewhere.
> Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ipxe.org/pipermail/ipxe-devel/attachments/20131101/29a68f7d/attachment.htm>

More information about the ipxe-devel mailing list