This is the mail archive of the
ecos-bugs@sourceware.org
mailing list for the eCos project.
[Bug 1001177] Redboot DHCP client race condition, XID, and retryproblems.
- From: bugzilla-daemon at bugs dot ecos dot sourceware dot org
- To: ecos-bugs at ecos dot sourceware dot org
- Date: Thu, 24 Mar 2011 18:12:41 +0000
- Subject: [Bug 1001177] Redboot DHCP client race condition, XID, and retryproblems.
- Auto-submitted: auto-generated
- References: <bug-1001177-13@http.bugs.ecos.sourceware.org/>
Please do not reply to this email. Use the web interface provided at:
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001177
--- Comment #3 from Grant Edwards <grant.b.edwards@gmail.com> 2011-03-24 18:12:39 GMT ---
> - in changelog "fixrace" -> "fix race". Also reference this issue number.
Done.
> - It would be nice (although admittedly arguably an enhancement) for
> all of bootp/dhcp support to be removable by config option. There
> are many instances where redboot is only ever given a static IP,
> so this code is just taking up space.
Isn't that what CYGSEM_REDBOOT_DEFAULT_NO_BOOTP does? I didn't mess
with the logic that controls what gets built and what doesn't, I was
just attempting to fix the DHCP support. I assumed that setting
CYGSEM_REDBOOT_DEFAULT_NO_BOOTP would build without BOOTP support.
...
But, I see that's not what it does. It just changes the default value
for a run-time flag that determines whether the call is made to
__bootp_find_local_ip().
> If that happens, obviously a few existing CDL config options would then live
> under that CDL component.
>
> I notice (in existing code, not your patch) that
> CYGSEM_REDBOOT_NETWORKING_USE_GATEWAY depends on
> CYGSEM_REDBOOT_NETWORKING_DHCP, which seems spurious to me.
> - In various places in the patch, constants like 4 or 6 are
> used. While entirely accurate, and they'll never change, from the
> principle of self-documenting code it would be nicer if they were
> things like e.g. sizeof(b->bp_chaddr) or
> sizeof(__local_enet_addr) etc. I admit I'm being a bit idealistic.
Done.
> - It's not clear to me in bootp_handler that this supports BOOTP
> even when CYGSEM_REDBOOT_NETWORKING_DHCP is enabled. Can you
> confirm it does?
Off the top of my head, I don't think it does.
> Just because CYGSEM_REDBOOT_NETWORKING_DHCP is
> enabled, shouldn't mean that we can't work with BOOTP any
> more. This may have been a problem in the previous code too.
I don't know. I've never tried it with a server that doesn't do DHCP.
Will a bootp-only server send a BOOTP RESPONSE in reply to a DHCP
DISCOVER?
I'll look into it.
> - You now depend on the CRC code for xid generation. That's quite a chunky
> dependency to introduce, and slightly pointless given there's very little
> randomness from every byte of the MAC address.
OK, I'll get rid of the call to the CRC routine.
> I think it would be good enough to just take the lower 4 bytes of
> the MAC (or if you prefer, take the bottom four, and then XOR in
> the top 2). You could maybe XOR in MS_TICKS()
I tried calling HAL_CLOCK_READ(), but the value was always 1. I'll
try MS_TICKS();
> for a little more entropy, although not much I agree. Or for a bit
> more, XOR in &xid and &__bootp_find_local_ip thus making it more
> specific to the specific build of the executable.
That sounds like a good idea
> I know that CYGPKG_CRC comes from the template, but that doesn't
> mean the code is included - only if it's used. Admittedly we should
> also add on explicit CDL dependencies where it's already been used -
> it's rather lax that that hasn't been done for
> CYGBLD_BUILD_REDBOOT_WITH_CKSUM and CYGSEM_REDBOOT_FIS_CRC_CHECK.
I'll skip the CRC call.
> Formally, I'll mark this patch as not having passed review, even
> though the changes are minor. Also please don't set the assignment
> flag until we hear from the FSF. I still haven't had a reply from
> them. I have no doubt you filled in the assignment, but we've had
> things before when a company has made changes to the assignment
> text, or changes to the copyright disclaimer, or has omitted the
> disclaimer entirely in what they sent. So I want to allow for the
> possibility that something goes wrong. The copyright is only
> actually assigned once the FSF executes the assignment on their side
> too, so until then, the copyright still rests with you, even if
> you've signed.
OK.
--
Configure bugmail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.