[#CORE-2272] Server starts returning garbage when killing events connection attempt

Firebird Core

Server starts returning garbage when killing events connection attempt

Created: 11/Jan/09 11:17 AM Updated: 12/Nov/09 02:14 PM

Component/s:

Engine

Affects Version/s:

2.1.1

Fix Version/s:

2.5 Beta 2, 2.1.3, 2.0.6

Time Tracking:

Not Specified

File Attachments:

CORE-2272.7z (543 kB)
2.

test.zip (184 kB)

Environment:

FB 2.1.1 SS, Linux 32b NPTL

Issue Links:

Relate

This issue relate to:

~~CORE-2437~~ Buffer overflow on client when delivering events.

Planning Status:

Unspecified

Description

« Hide

1. I configured my FB2.1 SuperServer events port to 32768.
2. I've connected to it. So far so good. In response right IP, right port.
3. I've connected again and between port reading and address reading
from response, I've killed the program.
4. From now, I'm getting back weird IPs and weird ports - probably
some numbers from wrong memory location.

All

Comments

Work Log

Change History

Version Control

Subversion Commits

Sort Order:

[ Permalink | « Hide ]

Jiri Cincura added a comment - 11/Jan/09 11:19 AM

Run the application:
app.exe "hostname/port:database" "username" "password"

First attempt will be successful (right IP and port), second returns random data.

[ Permalink | « Hide ]

Alexander Peshkov added a comment - 12/Jan/09 04:57 AM

Can you send me source code for it?
I can't start .Net under wine - it fails in x11 driver.

[ Permalink | « Hide ]

Jiri Cincura added a comment - 15/Jan/09 04:28 PM

Tweaked sources with test app. are attached.

[ Permalink | « Hide ]

Alexander Peshkov added a comment - 13/Mar/09 11:05 AM

Jiri, I confirm the issue. But not only port number and IP is bad - the whole event delivery in second attach is broken, I get AVs, assertions, etc. Will investigate more.

[ Permalink | « Hide ]

Alexander Peshkov added a comment - 17/Apr/09 06:45 AM

Jiri, after fixing http://tracker.firebirdsql.org/browse/CORE-2437 I can't reproduce an issue. Without fixing also could not, but due to getting segfaults :))

Can you check it with HEAD? There were BOFs before fix (and they happened exactly when I tried to connect second time). Therefore I suppose current issue may be also gone now. Though problem may be also due to .Net client (I fixed only fbclient).

I anyway plan to backport 2437 to 2.1 and 2.0, but wish to have your results first.