German mirror provided by / Deutscher Spiegelserver von  &  SEO LEO

PuTTY bug portfwd-corrupt

Home | FAQ | Feedback | Licence | Updates | Mirrors | Keys | Links | Team
Download: Stable · Snapshot | Docs | Privacy | Changes | Wishlist

We've had several reports of data corruption in port-forwarded connections. Some (possibly all) of these are due to data loss. The PuTTY client (as opposed to others) appears to be correlated with the data loss, although it's unclear whether the data is actually being lost before or after PuTTY is dealing with it.

Problems have been reported with both directions of tunneling, both directions of data transfer, and both SSH protocols (SSH-1 and SSH-2).

Update, 2005-06-15: I (JTN) have reproduced loss of trailing data with a remote-to-local SSH-2 port forwarding, where a burst data is sent from remote to local followed immediately by EOF. I suspect that PuTTY's lack of support for half-closed connections is to blame in at least this case. Update, 2011-09-13: I (SGT) have just (hopefully) fixed the half-closed issue, so perhaps that will be improved now.

The reports below may correspond to several different bugs, as a range of symptoms are reported.

• 2821c598.0306222343.8ea8a79@posting.google.com
  (also 3F031766.BC09B85A@uni-mb.si)
  Server: SSH-1.5-OSU_1.4alpha3 on OpenVMS V7.3-1
  Protocol: SSH-1
  local-to-remote tunnel of POP3/SMTP
  Problems with sending messages (dialogue or attachments corrupted)
  Apparently the first port-forwarded connection in a session will lose some quantity of the data sent from PuTTY to the server from the start of the connection.
  SSH packet log makes it look like data is being lost by the server (or possibly after PuTTY does the logging), but other clients (e.g., OpenSSH, recent F-Secure) don't show this behaviour.
• 20040209143157.89534.qmail@web14102.mail.yahoo.com
  0.53b + (2004-02-02?) + (2004-02-09?)
  Various Red Hat 7.x servers. One is SSH-1.99-OpenSSH_3.1p1.
  Protocol: SSH-1 and SSH-2
  HTTP tunnel (local-to-remote?)
  Result of GET truncated (incomplete)?
  Looks to be correlated with slow consumer (web browser)
  Same person has same problem with TeraTerm SSH ("Error 10058: Cannot send after socket shutdown.").
  PuTTY sometimes hangs?
  Includes SSH packet log; this shows SSH-2 local-to-remote forwarding, with PuTTY apparently receiving all the data. Maybe it fails to pass it on to the HTTP client?
• 002101c45ee8$ff761990$6801a8c0@carmani600m
  "SMTP tunneling corrupts attachments"
  It looks like something other than PuTTY was modifying the attachments (possibly in addition to PuTTY), so we aren't able to extract useful information from this report.
• 1090249727.4682.55.camel@oberon.tremagi.org.uk
  Win2KSP3 + PuTTY 0.54 (and 0.53b)
  Fedora Core 2 (unpatched) + OpenSSH (openssh-server-3.6.1p2-34)
  Remote tunnel for "yum" over HTTP: "yum" pulling data (i.e. outbound from PuTTY)
• D9304F864F1759059287F365@[192.168.0.10]
  0.53b, 0.58, 2005-06-12:r5952 to OpenSSH 3.9p1
  Using SOCKS dynamic forwarding to forward FTP, upload corrupts files (resulting file is shorter than expected); FTP progress bar goes suspiciously fast for first 1-200k of a 500k file
  Other SOCKS servers behave fine; `get' is fine; static port forwarding is fine
• 6.0.0.22.2.20050614193231.01fda408@195.70.37.56
  0.58 to Debian OpenSSH
  Remote-to-local tunnel for JetDirect (local printer port 9100)
  Files under 8k: OK
  Files over 8k: all data in PuTTY's log, but not all gets to the printer
  Logs available

SGT, 2024-11-17: classifying this bug as historic. It's too vague to investigate properly: just "corruption" is unclear about exactly what happens. The only known problem was truncation of the end of the data in cases of unidirectional EOF, which as far as we know was fixed in 2011 by changes to EOF handling.