nfsd in the Linux kernel before 2.6.28.9 does not drop the CAPMKNOD capability before handling a user request in a thread, which allows local users to create device nodes, as demonstrated on a filesystem that has been exported with the rootsquash option.
Created attachment 331637 [details] ipdstcache (in blue) VS route -Cn | wc -l (in green) over time
Description of problem: The value of ipdstcache (in /proc/slabinfo) continues to grow constantly, even thought the cached route table remains fairly constant. This leads to the eventual time when ipdstcache reaches the value of /proc/sys/net/ipv4/route/maxsize. When this happens, the kernel complains with 'dst cache overflow' and the server no longer responds to any network activity.
Version-Release number of selected component (if applicable): kernel 2.6.18-92.1.22.el5
How reproducible: live system being currently affected by this issue.
Steps to Reproduce: 1. Configure test machine as a router between two networks 2. send packets from network A to network B with a large number of different source/dest IPs 3. Watch the values of ipdstcache and rtcache Actual results: ipdstcache continues to grow while rtcache grows and shrinks with the traffic
Expected results: ipdstcache and rtcache follow each other closely. Values return to zero after traffic stops and route cache entries expire.
Additional info:
Multiple use-after-free vulnerabilities in libxml2 2.5.10, 2.6.16, 2.6.26, 2.6.27, and 2.6.32, and libxml 1.8.17, allow context-dependent attackers to cause a denial of service (application crash) via crafted (1) Notation or (2) Enumeration attribute types in an XML file, as demonstrated by the Codenomicon XML fuzzing framework.
net/unix/afunix.c in the Linux kernel 2.6.31.4 and earlier allows local users to cause a denial of service (system hang) by creating an abstract-namespace AFUNIX listening socket, performing a shutdown operation on this socket, and then performing a series of connect operations to this socket.
Multiple race conditions in fs/pipe.c in the Linux kernel before 2.6.32-rc6 allow local users to cause a denial of service (NULL pointer dereference and system crash) or gain privileges by attempting to open an anonymous pipe via a /proc//fd/ pathname.
The execve function in the Linux kernel, possibly 2.6.30-rc6 and earlier, does not properly clear the current->clearchildtid pointer, which allows local users to cause a denial of service (memory corruption) or possibly gain privileges via a clone system call with CLONECHILDSETTID or CLONECHILDCLEARTID enabled, which is not properly handled during thread creation and exit.
Untrusted search path vulnerability in VMware vMA 4.x and 5.x before 5.0.0.2 allows local users to gain privileges via a Trojan horse DLL in the current working directory.