(Note: these changelogs are out of date, I put some info for later releases in the NEWS page; also, if you download the splitout version of the patches, you'll find detailed patch descriptions in each of them).

Guest patches

2.6-bb tree

2.6.9-bb4:

(Now should be really) fixed tons of zombie threads created when running in TT mode under a 2.6.9 host.
Some stability issues were fixed - uml-extramask is all about this.

+uml-fix-sigreturn-atomicity.patch

Fix a damn warning about code atomicity (probably wasn't biting anyone of you -
you'd need a SMP UML to be bitten by this).

-uml-rework-uml-hang-on-2.6.9-host.patch

drop that because not good

+uml-hang-on-2.6.9-host.patch

Readded my original one

+uml-jdike-fix-tt-exit-2.6.9-host.patch

Added this from Jeff incrementals... it should be the latest version (let me now
in case I'm wrong.

+uml-extramask.patch

Fixes a big stability issue, with realtime signal delivery not working well and
causing MySQL to generate tons of zombie threads.

+uml-timer-context.patch

Fix for a check_rlimit bug (this actually won't probably hit you in this kernel
revision - here is a cleanup; the bug it fixes can be triggered only in
2.6.10-rc2 due to some mainline changes).

2.6.9-bb3:
- (Probably) fixed tons of zombie threads created when running in TT mode under a 2.6.9 host.
- Fixed compilation on Slackware 10 and other systems having 2.4 /usr/include/{linux,asm} kernel headers installed.
- Partially fixed the reboot problem (works for Bodo Stroesser).
- Dropped for now DevFS changes (uml-remove-devfs-mk-symlink.patch) mentioned in the FAQ.
- Dropped some little improvements in startup checks for presence of SKAS support, because there is one unconfirmed report they are causing problems.
- Remove some code which was duplicated by mistake in previous releases.
- Disable SYSEMU support in TT mode because it has some security problems and not a lot of performance advantages. (uml-disable-sysemu-when-tt.patch).
```
+uml-fix-sysemu-tt.patch

Fix SYSEMU and TT mode interaction, probably (not sure, so I disabled the
combination in uml-disable-sysemu-when-tt.patch).

+uml-fix-reboot-skas.patch

Fix reboot in SKAS mode

+uml-close-all-fds.patch

Close all open file descriptors on reboot to avoid UML failing the reboot.

+uml-rework-uml-hang-on-2.6.9-host.patch

Bodo's version of the fix to make UML run on a 2.6.9 host.

    
```
2.6.9-bb2:
- Fixed the bug when compiling without CONFIG_MODE_TT, which made it fail on startup; that bug was introduced in -bb1.
- Fixed the bug on Gentoo (and any host system having only a NPTL glibc) which caused UML not to startup, when compiling only SKAS mode. This existed since 2.6.0, and turned out to be a bug in glibc. If you enable CONFIG_MODE_TT, however, it still refuses to link :-(, and fixing that is harder.
- Dropped uml-use-SIG_IGN-for-empty-handling.patch, which was introduced in -bb1 and was wrong. This probably caused processes inside UML consoles not to notice xterm resizes.
2.6.9-bb1:
- Runs fine (without hanging on exit) on 2.6.9 hosts.
  Actually, it was later discovered that this is true only for SKAS mode (see 2.6.9-bb3).
- Included a lot of security fixes from Bodo Stroesser
- Lots of little changes.

2.4-bs tree

2.4.27-bs1:
- Runs fine (without hanging on exit) on 2.6.9 hosts.
  (Later note): I think this is only true for SKAS mode, since I've included the same fix as in 2.6.9-bb1.
- The hostfs rewrite started after 2.4.24-1um is not included, so hostfs works well.
- Included safe fixes from the patches after 2.4.24-1um, including incrementals, except for SYSEMU.
- Diagnosed the "hwclock hangup" in TT mode when using Debian-3.0r2 rootfs. If you experience it, revert the SIG-filehandle-changes-part-2.patch and SIG-fileHandle-changes.patch patches (you'll find them in the broken-out version). But without those, the kernel will not be stable on 2.6 hosts in TT mode.
  SKAS mode is unaffected by all these problems, so using it is the simpler solution.

Host patches

2.6 host patches

Skas3 + SYSEMU patch

Here is the changelog, starting from 2.6-V1.

2.6-V8-rc2 (released on 11 February) :
- Released against 2.6.7, 2.6.9, 2.6.10 and 2.6.11-rc3-bk6. The 2.6.7 one is ok on 2.6.8.1.
- for kernels >=2.6.10, in the splitout, various patches were folded together.
- rename-modify-ldt-to-avoid-conflicts.patch: fix the reported ATI module compile problem by renaming modify_ldt to __modify_ldt.
- ptrace_ldt-reload-smp-fix.patch: fix a SMP-host bug noticed by Bodo Stroesser.
- skas-add-locking-for-mm-switch.patch: from me, add correct locking to PTRACE_SWITCH_MM to avoid races. Someone please test with SMP hosts / with spinlock debugging enabled, to check for any possible deadlock. I'll do this in next days, too.
- skas-reorganize and skas-add-SYSEMU_SINGLESTEP: from Bodo Stroesser, they add the SYSEMU_SINGLESTEP mode. For 2.6.9 and beyond, I've cleaned up a lot the second one to ease porting to 2.6.10.
- sysaudit-singlestep-umlhost.patch: for host >= 2.6.9, it fixes a bad interaction between syscall auditing and ptracing. A version of this must also go in 2.6.11.
- fix-dumpable-handling.patch: this implemented a feature / bug fix requested by Michael Richardson (mcr (at) sandelman (dot) ottawa (dot) on (dot) ca). I.e., with this, in normal cases, UML *is* allowed to dump core and be seen inside /proc - (until now it was treated as if it was a setuid process or one which changed UID / GID). If I did any error, however, it could allow privilege escalation. Well, I've thought a bit about this, however I'll think even more before putting it into a release (I wrote this more than one month ago).
2.6-V7 (released on 29 October) :
- Released against 2.6.7 and 2.6.9. The 2.6.7 one is ok on 2.6.8.1.
- Bodo Stroesser identified and fixed some other problems with SYSEMU support in the host patch. Now
```
echo 0 > /proc/sysemu
```
  works well, as it does on 2.4 hosts.
- Also fixed in the same way the support for singlestepping inside UML.
- For 2.6.7 release only: thanks to a Gerd Knorr suggestion, an important ptrace bugfix from 2.6.9 has been backported to 2.6.7.
- Add -skas3-v7 to the kernel extraversion only on 2.6.9, by creating localversion-skas (using a new Kbuild feature).
- No 2.6.8.1 release: the 2.6.7 is ok this time, I had to avoid changing EXTRAVERSION. You can do this by hand at will, if you like so.
2.6-V6 (released on 18 October) :
- Bodo Stroesser identified and fixed some problems with SYSEMU support, of which ones was in the host patch. Actually, this causes problems if after calling ptrace(PTRACE_SYSEMU,...) UML calls ptrace(PTRACE_SYSCALL,...) on the same child process. This makes some releases of UML (the 2.4 ones with Jeff Dike's version of SYSEMU support) panic on startup.
- For 2.6.9-rc4 release only: Gerd Knorr identified an issue which needed some updates (see in mm/proc_mm.c the call to arch_pick_mmap_layout(), which is new in 2.6.9-rc2).
- Add -skas3-v6 to the kernel extraversion. On 2.6.9 this is done by creating localversion-skas (a new Kbuild feature).
- Release a separate patch for 2.6.8.1: the only reason is that the patch on the Makefile fails against 2.6.8.1.
2.6-V5 (released on 19 September) : fix a silly error (by me) which could create problems when using a debugger (never seen it, can be depending on gcc version). Actually it's a "return;" instead of a "return 0;". Noticed by Gerd Knorr < kraxel (at) bytesex (dot) org >.
2.6-V4 (released on 12 September) : make SYSEMU work even with SYSENTER (this does not interact with UML, but with other code) + clean up some more details about syscall entry code in entry.S.
2.6-V3: fixes a deadlock which can be triggered when being low on memory; also upgrade the SYSEMU patch due to some ideas by Jeff Dike, to make it faster and cleaner, and fix some little bugs with kernel APIs
2.6-V2: some people (including me) had a special setup which made the host kernel leak tons of memory. Fix included. Also, it contains fixes to the mprotect(2) support for UML, and the SYSEMU patch from Laurent Vivier.
2.6-V1: after various prereleases, and discussions, a bug on SMP was discovered, leading to release of 2.6-V1 containing the fix for SMP systems. SYSEMU is not included.

2.4 host patches

Skas3 + SYSEMU patch

2.4-V3: fixes a deadlock which can be triggered when low on memory.
2.4-V2: backport the changes in 2.6 V2 release (including SYSEMU).
2.4-V1: port of the original "host-skas3.patch" by Jeff Dike to the 2.4.25 kernel; the original one became incompatible because in 2.4.25 a rewrite of LDT handling (arch/i386/kernel/ldt.c) was backported from 2.6. SYSEMU is not included.