Hi there, I'm experiencing a strange, annoying ssh daemon problem, I'm rather desperated: At transmitting huge files (about 2GB) via scp to the ssh server, I earn an "Connection reset by peer" in about 90% of the cases, in average between the 100th and 1800th megabyte of transfer. Long-lasting ssh sessions without huge data amounts remain untouched. Tracking down the issue, I found out that the TCP session is terminated by flags RST,ACK, for which RFC793 says "connection not known to the stack". Due to this, I started verbose logging of the daemon, the end of the log created by sshd -ddd follows here: --------------------cut--------------------- debug2: channel 0: window 61440 sent adjust 69632 <many of them> debug2: channel 0: window 61440 sent adjust 69632 debug2: channel 0: window 61440 sent adjust 69632 Segmentation fault --------------------cut--------------------- Nothing further, of course the process is dead now. The client tells me the following: --------------------cut--------------------- <usual connection-rising-crap> debug3: channel 0: close_fds r -1 w -1 e -1 test.zero |========> 75% 1509MB 10.5MB/s 00:46 ETA Read from remote host svdbslx007: Connection reset by peer lost connection --------------------cut--------------------- I see, referencing to the received RST the client could not do anything else :) The SSH server is a SLES9 32bit, I already updated to SP1 and SP2 - no change of the behaviour. Actually the sshd version is OpenSSH_3.8p1. As client I tried several SuSE versions (9.3 Prof at current patch level, SLES9 32bit at SP2, SLES9 64bit SP1, etc...) and my Debian Sarge stable notebook. Always the same. Are there any suggestions? Thanks in advance and best regards, -- Mit freundlichen Grüßen i.A. Martin Klier PC-Benutzerunterstützung / Linux-Server IT Asset Management A.T.U - Auto-Teile-Unger Dr.-Kilian-Straße 11 92637 Weiden
Attachment:
pgpSpaZhnFOdZ.pgp
Description: signature
