Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Dirk Wetter
Package: mktemp Version: 1.5-2 Severity: grave Tags: security There's a problem with the randomness of mktemp. The string includes a number which includes somewhat the current process ID (based on the current PID). Worse: Subsequent calls just seem to increase the number by one: [EMAIL

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Nico Golde
severity critical thanks Hi Dirk, * Dirk Wetter [EMAIL PROTECTED] [2008-08-15 11:57]: There's a problem with the randomness of mktemp. The string includes a number which includes somewhat the current process ID (based on the current PID). Worse: Subsequent calls just seem to increase the

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Clint Adams
Possible solutions: 1) change the algorithm 2) use libc mkstemp (will break templates shorter than 6 characters) 3) ship mktemp from sarge (will break templates shorter than 6 characters) 4) ship coreutils mktemp (will break templates shorter than 3 characters) -- To UNSUBSCRIBE, email to

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Sven Joachim
I raised the severity to critical as a lot of shell scripts in Debian rely on this data being random. Why is that important? The purpose of mktemp is to return a unique filename and to actually create the file. Can you describe an attack based on the non-randomness of the filename? Sven

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Dirk Wetter
Am 15.08.2008 14:02, Clint Adams schrieb: Possible solutions: 1) change the algorithm 2) use libc mkstemp (will break templates shorter than 6 characters) 3) ship mktemp from sarge (will break templates shorter than 6 characters) 4) ship coreutils mktemp (will break templates shorter than 3

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Andreas Metzler
On 2008-08-15 Clint Adams [EMAIL PROTECTED] wrote: Possible solutions: 1) change the algorithm 2) use libc mkstemp (will break templates shorter than 6 characters) 3) ship mktemp from sarge (will break templates shorter than 6 characters) 4) ship coreutils mktemp (will break templates

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Dirk Wetter
Am 15.08.2008 14:42, Andreas Metzler schrieb: On 2008-08-15 Clint Adams [EMAIL PROTECTED] wrote: is there actually a reason for using the PID and for limiting the non-pid part to [[A-Za-z]? Why not bugs included- --- mktemp-1.5.orig/priv_mktemp.c +++ mktemp-1.5/priv_mktemp.c @@

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Nico Golde
Hi Sven, * Sven Joachim [EMAIL PROTECTED] [2008-08-15 15:40]: I raised the severity to critical as a lot of shell scripts in Debian rely on this data being random. Why is that important? The purpose of mktemp is to return a unique filename and to actually create the file. Can you

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Sven Joachim
On 2008-08-15 14:42 +0200, Andreas Metzler wrote: is there actually a reason for using the PID and for limiting the non-pid part to [[A-Za-z]? Why not bugs included- --- mktemp-1.5.orig/priv_mktemp.c +++ mktemp-1.5/priv_mktemp.c @@ -81,25 +81,20 @@ register char *start,

Bug#495193: Etch's and Lenny's mktemp seriously broken

2008-08-15 Thread Andreas Metzler
On 2008-08-15 Sven Joachim [EMAIL PROTECTED] wrote: On 2008-08-15 14:42 +0200, Andreas Metzler wrote: is there actually a reason for using the PID and for limiting the non-pid part to [[A-Za-z]? [...] - char *alphabet = ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz; + char