Skip to contentSkip to conversation

EAGAIN and Linux 2.6 IPsec patch

DISCUSSION dual_bereta_r0x 1 participant 1 message Mar 23, 2004

lore
Share & tools
lore
#0 lore raw

Is this fixed already? Please cc me too.

On Tue, 30 Sep 2003 15:41:51 +0400
Alexey <[email protected]> wrote:

Hello!

Quoted 1 line

The following (untested) patch tries to do that.

This does not work and has no chances to do. We have already discussed this, actually. You can get reasoning from saved mailbox, I think. To resume, proto->connect() method must not ever block.

Alexey

PS. I think I am back to life, so this time I am honest telling
this bug is going to be fixed properly.
-
FreeS/WAN design list.
https://mj2.freeswan.org/cgi-bin/mj_wwwusr/domain=mj2.freeswan.org to unsubscribe

Quoted 10 lines · On Tue, 30 Sep 2003 21:34:24 +1000 Herbert Xu <[email protected]>…

OK, this looks like a bug that is independent of the packet queueing issue.

TCP/UDP connects do a route/IPSEC lookup to determine information about the connection. The IPSEC lookup can fail with EAGAIN if the policy is in place without any SAs.

Since connect already has an O_NONBLOCK bit we should make it try harder if it is clear.

The following (untested) patch tries to do that.

Please don't do this, any attempt to fix the EAGAIN behavior with hacks like this are riddled with problems.

Several patches nearly identical to your's have been submitted in the past.

Alexey is supposed to work on support for unresolved routes, which is the proper way to fix this problem.
-
FreeS/WAN design list.
https://mj2.freeswan.org/cgi-bin/mj_wwwusr/domain=mj2.freeswan.org to unsubscribe

Signature · 2 lines
dual_bereta_r0x -- Alexandre Hautequest
ArenaNetwork Lan House & Cyber -- www.arenanetwork.com.br