Mark Huang [Tue, 1 Feb 2005 03:19:52 +0000 (03:19 +0000)]
- bound vnet mods by #ifdef
- bound runtime vnet code by if (vnet_active)
Mark Huang [Tue, 1 Feb 2005 03:15:09 +0000 (03:15 +0000)]
- bound vnet mods by #ifdef
- bound runtime vnet code by if (vnet_active)
- vnet needs inet_family_ops, export if configured
- PL3976 fix: inet_release(): don't reset sk_xid
Mark Huang [Tue, 1 Feb 2005 03:12:33 +0000 (03:12 +0000)]
- bound vnet mods by #ifdef
Mark Huang [Tue, 1 Feb 2005 03:11:34 +0000 (03:11 +0000)]
- vnet needs dev_queue_xmit_nit(), export it
- bound vnet mods by #ifdef
Mark Huang [Tue, 1 Feb 2005 03:10:21 +0000 (03:10 +0000)]
- bound vnet mods by #ifdef
- bound runtime vnet code by if (vnet_active)
Mark Huang [Mon, 31 Jan 2005 20:33:28 +0000 (20:33 +0000)]
fix inline declaration, gcc-3.4 complains
Marc Fiuczynski [Fri, 28 Jan 2005 05:16:37 +0000 (05:16 +0000)]
fixed merge mistake that picked up 2.6.9 rather than 2.6.9-1.11_FC2 code
Marc Fiuczynski [Fri, 28 Jan 2005 05:14:00 +0000 (05:14 +0000)]
cleanup to reduce diff w/ 2.6.9-1.11_FC2
Marc Fiuczynski [Fri, 28 Jan 2005 05:13:32 +0000 (05:13 +0000)]
sync up with 2.6.9-1.11_FC2 release
Marc Fiuczynski [Fri, 28 Jan 2005 05:12:43 +0000 (05:12 +0000)]
fixed merge bug
Marc Fiuczynski [Fri, 28 Jan 2005 05:09:30 +0000 (05:09 +0000)]
moved stackoverflow checking out of mainline files
Andy Bavier [Thu, 27 Jan 2005 22:45:17 +0000 (22:45 +0000)]
Comment out return statement that prevents shares from being initialized correctly
Marc Fiuczynski [Fri, 21 Jan 2005 05:04:36 +0000 (05:04 +0000)]
upgrade to linux 2.6.9-1.11_FC2
Marc Fiuczynski [Fri, 21 Jan 2005 03:41:42 +0000 (03:41 +0000)]
Initial revision
Marc Fiuczynski [Fri, 21 Jan 2005 03:34:18 +0000 (03:34 +0000)]
This commit was generated by cvs2svn to compensate for changes in r366,
which included commits to RCS files with non-trunk default branches.
Marc Fiuczynski [Fri, 21 Jan 2005 03:34:16 +0000 (03:34 +0000)]
This commit was generated by cvs2svn to compensate for changes in r363,
which included commits to RCS files with non-trunk default branches.
Mark Huang [Fri, 21 Jan 2005 03:33:58 +0000 (03:33 +0000)]
This commit was generated by cvs2svn to compensate for changes in r344,
which included commits to RCS files with non-trunk default branches.
Marc Fiuczynski [Fri, 21 Jan 2005 03:33:53 +0000 (03:33 +0000)]
This commit was generated by cvs2svn to compensate for changes in r338,
which included commits to RCS files with non-trunk default branches.
Marc Fiuczynski [Wed, 19 Jan 2005 22:42:27 +0000 (22:42 +0000)]
Fixes CAN-2004 security reports 1016, 1017, and 1068
Mark Huang [Mon, 17 Jan 2005 20:07:16 +0000 (20:07 +0000)]
- patch from https://bugzilla.netfilter.org/bugzilla/show_bug.cgi?id=40
fixes crash if PPTP packets are received out of order
- PPTP headers are variable length. Specifically, PPTP ACKs without
sequence numbers are 4 bytes shorter and will not be properly
tracked. Pass the minimum length required to skb_header_pointer() when
trying to parse a generic PPTP header out of a packet.
- Fix a call to DUMP_TUPLE_GRE()
Mark Huang [Mon, 17 Jan 2005 20:02:29 +0000 (20:02 +0000)]
- PL3791 fix: Suppress RST if the port was bound to a (presumably raw) socket
Mark Huang [Mon, 17 Jan 2005 20:01:59 +0000 (20:01 +0000)]
- PL3791 fix: Suppress ICMP Unreachable if the port was bound to a (presumably raw) socket
Marc Fiuczynski [Fri, 14 Jan 2005 21:24:42 +0000 (21:24 +0000)]
Added some debug printk's for shares_write problems that Steve is seeing
on a bunch of nodes. It is not clear what exactly is causing the problem
he (i.e., NM) is observing. Hopefully these debug messages will reveal
something.
Marc Fiuczynski [Tue, 11 Jan 2005 04:56:07 +0000 (04:56 +0000)]
There is a bug in the CKRM CPU scheduler. This has been reported to the
folks at IBM. I am going to back out of the scheduler changes until I
am certain that the scheduler works as well as what we have now (ideally
better).
Mark Huang [Mon, 10 Jan 2005 20:31:05 +0000 (20:31 +0000)]
compile PPTP/GRE connection tracking modules
Marc Fiuczynski [Mon, 10 Jan 2005 19:43:12 +0000 (19:43 +0000)]
CKRM CPU controller version 9 rc1.
Mark Huang [Mon, 10 Jan 2005 19:31:58 +0000 (19:31 +0000)]
- backport 2.6.9 addition of skb_header_pointer() macro to support
pptp-conntrack-nat patch
Mark Huang [Mon, 10 Jan 2005 19:06:38 +0000 (19:06 +0000)]
- fix DEBUGP() macro; newer gcc versions refuse to concatenate __FILE__
and __FUNCTION__
- fix DUMP_TUPLE_GRE() macro
- 2.6.9 changed print_tuple() and print_conntrack() methods to use
seq_printf(), but we are running 2.6.8, so change to just sprintf()
Mark Huang [Mon, 10 Jan 2005 19:05:12 +0000 (19:05 +0000)]
- fix DEBUGP() macro; newer gcc versions refuse to concatenate __FILE__
and __FUNCTION__
- exp_gre: expectations must now be allocated by the helper/caller; it's
been this way for at least a year, this patch was probably never
tested
Marc Fiuczynski [Mon, 10 Jan 2005 19:02:43 +0000 (19:02 +0000)]
Initial revision
Mark Huang [Mon, 10 Jan 2005 19:02:29 +0000 (19:02 +0000)]
patch-o-matic-ng-
20050103 pptp-conntrack-nat
Mark Huang [Mon, 10 Jan 2005 18:58:16 +0000 (18:58 +0000)]
patch-o-matic-
20050103 pptp-conntrack-nat
Mark Huang [Mon, 10 Jan 2005 18:55:08 +0000 (18:55 +0000)]
patch-o-matic-ng-
20050103 pptp-conntrack-nat
Mark Huang [Mon, 10 Jan 2005 18:38:07 +0000 (18:38 +0000)]
patch-o-matic-ng-
20050103 conntrack-acct
Mark Huang [Mon, 10 Jan 2005 18:31:25 +0000 (18:31 +0000)]
patch-o-matic-ng-
20050103 pptp-conntrack-nat conntrack-acct
Mark Huang [Mon, 10 Jan 2005 18:29:17 +0000 (18:29 +0000)]
patch-o-matic-ng-
20050103 pptp-conntrack-nat
Steve Muir [Fri, 7 Jan 2005 21:13:59 +0000 (21:13 +0000)]
Incorporate vserver 1.9.3 changes to check_kill_permission (eliminates
call to vx_check in group_send_sig_info)
Steve Muir [Fri, 7 Jan 2005 21:12:01 +0000 (21:12 +0000)]
Check that SIGALRM got delivered (should never fail)
Mark Huang [Tue, 4 Jan 2005 18:46:08 +0000 (18:46 +0000)]
- fix do_mprotect() prototype
Mark Huang [Wed, 29 Dec 2004 18:55:30 +0000 (18:55 +0000)]
- remove 2004.12.14 from EXTRAVERSION
Marc Fiuczynski [Tue, 28 Dec 2004 04:48:53 +0000 (04:48 +0000)]
Turn off CKRM IO controller. It should never have been turned on in the
first place until we are truly ready to deploy it.
Marc Fiuczynski [Mon, 27 Dec 2004 22:13:16 +0000 (22:13 +0000)]
ckrm e16 io controller merge v5
Marc Fiuczynski [Tue, 21 Dec 2004 21:50:42 +0000 (21:50 +0000)]
PL3009: Implemented a UDP variant of IPOD. Sending a packet to port 664
with the IPOD key as data will also cause the machine to restart.
Marc Fiuczynski [Fri, 17 Dec 2004 22:55:21 +0000 (22:55 +0000)]
Turns X86_STACK_CHECK on by default.
Marc Fiuczynski [Fri, 17 Dec 2004 20:41:32 +0000 (20:41 +0000)]
Updated stack overflow handling support.
Marc Fiuczynski [Thu, 16 Dec 2004 08:59:09 +0000 (08:59 +0000)]
Moved code around to fix bug related to ".previous" section definition.
Aaron Klingaman [Tue, 14 Dec 2004 19:47:57 +0000 (19:47 +0000)]
enable floppy module generation for boot cd
Aaron Klingaman [Tue, 14 Dec 2004 19:46:42 +0000 (19:46 +0000)]
enable kexec
Aaron Klingaman [Tue, 14 Dec 2004 19:35:17 +0000 (19:35 +0000)]
kexec patch backported from 2.6.10-mm
Aaron Klingaman [Tue, 14 Dec 2004 19:29:05 +0000 (19:29 +0000)]
accidently delete
Marc Fiuczynski [Tue, 14 Dec 2004 19:13:38 +0000 (19:13 +0000)]
This stack check implementation leverages the compiler's profiling (gcc -p)
support to determine precisely when there is a stack overflow. The compiler
generates a call to the special function 'mcount' in the prologue of each
function. This special 'mcount' function checks whether the stack has
grown beyond STACK_WARN size; and, if so, it will call the stack_overflowed
function, which generates a stack trace. If the stack has grown to STACK_PANIC
size, the stack_overflowed function will panic() the system.
The hope is that with this support we can track down precisely the execution
context when the stack overflows.
Marc Fiuczynski [Mon, 13 Dec 2004 22:00:04 +0000 (22:00 +0000)]
Added configuration support for:
1) 4KB, 8KB, and 16KB stacks
2) separate IRQSTACKS
3) stack overflow warnings at a user-specified size
The original FC code only supports 4KB kernel stacks w/ a separate,
per-cpu, IRQ stack and the stack overflow warning was printed whenever
the kernel approached 1/8 of the current stack size.
The new support permits one to select different stack sizes and
separate IRQ stack using configuration options. The default right
now is to use a 8KB kernel stack with a separate IRQ stack.
Mark Huang [Tue, 30 Nov 2004 16:32:22 +0000 (16:32 +0000)]
- return -EPERM instead of silently freeing
Marc Fiuczynski [Tue, 30 Nov 2004 15:59:12 +0000 (15:59 +0000)]
PL3069
Fixed memory leak of skbuff. Requires changes to accurately do statistics
and some additional cosmetic changes. But this should be sufficient for
the next release candidate.
Mark Huang [Tue, 23 Nov 2004 15:15:05 +0000 (15:15 +0000)]
PL3131 fix: prevent vservers from escaping chroot() barriers
Mark Huang [Tue, 23 Nov 2004 15:11:39 +0000 (15:11 +0000)]
- apply patch to enable legacy ioctl() programs such as 0.30 stable
util-vserver showattr/setattr to set the immulink and barrier
high bits
- this is patch-2.6.9-final-vs1.9.3-rc3-legacyimmulinkioctl.patch
From: Sam Vilain <sam@vilain.net>
Re: [Vserver] [PATCH] immulink ioctl is not available on vs1.9.3-rc2, even with CONFIG_VSERVER_LEGACY
To: vserver@list.linux-vserver.org
Date: 10/18/04 12:48 am
Mark Huang [Tue, 23 Nov 2004 15:08:22 +0000 (15:08 +0000)]
PL3131 fix: prevent vservers from escaping chroot() barriers
Mark Huang [Tue, 23 Nov 2004 15:08:03 +0000 (15:08 +0000)]
- make sure that all extended attributes are passed properly to the
inode structure
Mark Huang [Tue, 23 Nov 2004 15:07:40 +0000 (15:07 +0000)]
- apply ioctl() patch to make sure that only superuser or someone
capable of setting the immutable bit, can set the immulink bit
From: Sam Vilain <sam@vilain.net>
Re: [Vserver] [PATCH] immulink ioctl is not available on vs1.9.3-rc2, even with CONFIG_VSERVER_LEGACY
To: vserver@list.linux-vserver.org
Date: 10/19/04 12:17 am
Here's the missing link.
Steve Muir [Tue, 23 Nov 2004 15:00:18 +0000 (15:00 +0000)]
Fix comment
Mark Huang [Fri, 19 Nov 2004 19:08:15 +0000 (19:08 +0000)]
ignore generated junk
Steve Muir [Thu, 18 Nov 2004 20:06:31 +0000 (20:06 +0000)]
Need a way to manipulate vserver file attrs using a file descriptor
Steve Muir [Wed, 17 Nov 2004 18:59:31 +0000 (18:59 +0000)]
Remove XID check that prevents one vserver from reading files created in
another vserver but made visible by, say, a bind mount
Andy Bavier [Tue, 16 Nov 2004 19:41:55 +0000 (19:41 +0000)]
Fixed vserver memory accounting bug. Between v1.8 and v1.9, a call to vx_rsspages_inc() was lost; restored now.
Marc Fiuczynski [Tue, 9 Nov 2004 16:48:52 +0000 (16:48 +0000)]
Turned off CKRM memory controller.
Waiting for PL2899 to be resolved, which involves waiting for the
CKRM/IBM folks.
Mark Huang [Mon, 8 Nov 2004 16:07:55 +0000 (16:07 +0000)]
- PL292: bump RLIMIT_NOFILE (a.k.a. INR_OPEN) to 4096 from 1024
Mark Huang [Mon, 8 Nov 2004 15:51:17 +0000 (15:51 +0000)]
- 2.planetlab was branched, bump release number on the HEAD to
3.planetlab to avoid confusion. Branch release numbers will look like
2.1.planetlab, 2.2.planetlab, etc.
Marc Fiuczynski [Fri, 5 Nov 2004 09:56:50 +0000 (09:56 +0000)]
This patch fixes a crash problem in rbce when a class is deleted
Problem is that the as part of running the rules_list the class
is effectively deleted through put_class thus the value of cls->classtype
is undefined. Since on entry cls->classtype == classtype use
parameter classtype instead.
Hubertus Frankeh (frankeh@us.ibm.com)
Mark Huang [Thu, 4 Nov 2004 16:32:45 +0000 (16:32 +0000)]
bump release number
Marc Fiuczynski [Thu, 4 Nov 2004 12:18:26 +0000 (12:18 +0000)]
changed printk's to use KERN_DEBUG, KERN_ERR, or KERN_WARNING.
Mark Huang [Thu, 4 Nov 2004 04:18:20 +0000 (04:18 +0000)]
- build a placeholder kernel-vserver RPM that provides kernel and
kernel-drm so that rpm or yum inside a vserver does not have to
download and install a full, useless kernel package to satisfy tcpdump
dependencies
Mark Huang [Wed, 3 Nov 2004 18:48:24 +0000 (18:48 +0000)]
- PL2529 fix: disable vserver debug code
Mark Huang [Wed, 3 Nov 2004 18:47:17 +0000 (18:47 +0000)]
- don't install Kerntypes until we can fix LKCD
Mark Huang [Wed, 3 Nov 2004 18:08:32 +0000 (18:08 +0000)]
- disable LKCD until we can actually get it to work (i.e. update
patchset)
Marc Fiuczynski [Wed, 3 Nov 2004 11:49:53 +0000 (11:49 +0000)]
New documentation for CKRM.
Andy Bavier [Wed, 27 Oct 2004 21:09:03 +0000 (21:09 +0000)]
Commented out the 'interactive bonus' and 'cvt saving' when updating cvt
Andy Bavier [Wed, 27 Oct 2004 21:07:25 +0000 (21:07 +0000)]
Commented out the contribution of 'urgency' to a class's effective priority
Andy Bavier [Wed, 27 Oct 2004 21:06:11 +0000 (21:06 +0000)]
Increased size of CLASSQUEUE from 128 to 1024 to support CPU shares of 4 or greater
Steve Muir [Wed, 27 Oct 2004 20:12:10 +0000 (20:12 +0000)]
Permit hard-linking across bind-mounts as long as the source dir is not
read-only [PL #2624]
Mark Huang [Wed, 27 Oct 2004 15:09:37 +0000 (15:09 +0000)]
add miscellaneous tags
Marc Fiuczynski [Sat, 23 Oct 2004 20:13:55 +0000 (20:13 +0000)]
Disabled surplus allocation feature of the CKRM cpu scheduler.
See PL2642 for further details.
Marc Fiuczynski [Fri, 22 Oct 2004 06:26:45 +0000 (06:26 +0000)]
Increased CKRM_SHARE_ACCURACY from (1<<10) to (1<<13). This resolves PL2620.
Marc Fiuczynski [Thu, 21 Oct 2004 17:54:19 +0000 (17:54 +0000)]
Fix to prevent kernel from crashing in the memory controller on shutdown.
Mark Huang [Wed, 20 Oct 2004 02:33:41 +0000 (02:33 +0000)]
- PL2525 fix: implement SO_SETXID (reset the context ID of an existing
socket). Callable by root context only. To avoid having to register a
new sockopt number, #define SO_SETXID to be SO_PASSCRED (which is not
usually callable by setsockopt()).
Mark Huang [Wed, 20 Oct 2004 02:33:15 +0000 (02:33 +0000)]
- PL2525 fix: implement SO_SETXID (reset the context ID of an existing
socket). Callable by root context only. To avoid having to register a
new sockopt number, #define SO_SETXID to be SO_PASSCRED (which is not
usually callable by setsockopt()):
#if !defined(SO_SETXID)
#define SO_SETXID SO_PASSCRED
#endif
int xid;
setsockopt(fd, SOL_SOCKET, SO_SETXID, &xid, sizeof(xid));
Marc Fiuczynski [Tue, 19 Oct 2004 23:21:35 +0000 (23:21 +0000)]
have been renamed to ckrm_numtasks by the CKRM folks
Marc Fiuczynski [Tue, 19 Oct 2004 23:05:37 +0000 (23:05 +0000)]
Reenabled CKRM memory controller.
Marc Fiuczynski [Tue, 19 Oct 2004 22:55:10 +0000 (22:55 +0000)]
disabled code for CKRM memory class shrinking
Mark Huang [Mon, 18 Oct 2004 17:45:31 +0000 (17:45 +0000)]
- fix printing of conntrack owners
Marc Fiuczynski [Mon, 18 Oct 2004 13:24:29 +0000 (13:24 +0000)]
version 9.1 of the ckrm cpu scheduler.
Marc Fiuczynski [Fri, 15 Oct 2004 20:55:24 +0000 (20:55 +0000)]
commented out unnecessary BUG_ON statement
Marc Fiuczynski [Fri, 15 Oct 2004 20:53:04 +0000 (20:53 +0000)]
typo
Marc Fiuczynski [Fri, 15 Oct 2004 20:51:06 +0000 (20:51 +0000)]
enabled ckrm memory controller
Marc Fiuczynski [Mon, 11 Oct 2004 20:26:18 +0000 (20:26 +0000)]
Config file to be used for "kernel-desktop" rpm package. It includes a
bunch of stuff -- such as USB, cpufreq, and laptop support that we
currently do not wish to add to the standard
kernel-2.6.8-i686-planetlab.config file.
Mark Huang [Sun, 10 Oct 2004 22:24:53 +0000 (22:24 +0000)]
PL2340 workaround: disable CKRM memory controller
Mark Huang [Fri, 1 Oct 2004 20:10:17 +0000 (20:10 +0000)]
- initialize connection contexts to -1 (== unknown)
Mark Huang [Fri, 1 Oct 2004 20:09:49 +0000 (20:09 +0000)]
- Fix thinko. Bypass xid checks if socket was created in the root
context, like the freaking comment says.
Mark Huang [Fri, 1 Oct 2004 17:59:41 +0000 (17:59 +0000)]
- panic on OOM condition instead of killing processes
- compile hangcheck-timer into kernel
Mark Huang [Fri, 1 Oct 2004 17:55:04 +0000 (17:55 +0000)]
- reboot by default
Mark Huang [Fri, 1 Oct 2004 17:54:48 +0000 (17:54 +0000)]
- Just panic() instead of the default behavior of selecting processes
for death.