Messages in this thread | | | Date | Thu, 09 Dec 1999 18:28:09 -0800 | From | Zachary Amsden <> | Subject | Re: Windows 9x and RFC1323 [problems with tcp SACK option, help?] |
| |
It really wouldn't surprise me if one of the ends of your connection suffers from SACK brain damage. SACK can easily cause problems if not done very carefully. In a bi-directional data connection, you can have SACK options plus tcp data in the same packet, and if you don't adjust your sending data length to account for this, you get an over MTU packet, which is refused (or should be) by your device driver. In most unmodified BSD stacks, such an error is silently ignored. In addition, if your mbufs are small, the addition of TCP options can easily overflow the mbuf (on BSD stacks).
Since it requires some serious effort to stress all the possible SACK cases, it really wouldn't surprise me if some vendors have broken code out there.
Since the FTP-data connection isn't bi-directional, I would imagine it is the second problem you are hitting.
To help out broken bi-directional stacks, one can reduce the advertised window by 40 bytes.
I have a rather heavy duty testing setup which I can use to simulate just about any type of connection - I'll run some stress on the latest linux this weekend to see what shows up, and will verify or fix the linux end of things. Good luck with Winboxes though.
-- Zachary Amsden zamsden@engr.sgi.com (650) 933-6919 09U-510 Core Protocols
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |