Bug 8515

Hi Thomas !
I was quite ready to test this kernel ... but
To be installed, Kernel rt (or server or dektop...) needs kmod having been updated to version 7-7 from update_testing repo too...


kmod 7-7 itself has not been tested as an update...
If this update brings some problem  would there be a simple possibility to downgrade it to 7-6... or is there a risk to have to reinstall Mageia2 because no boot is anymore possible ? 

Wouldn't it be more safe to propose first kmod 7-7 as update request to QA to verify if it is strictly compatible with previous stable kernels ?

Just a thought .

Thanks for your huge work.

Philippe

CC: (none) => philippedidier

kmod is already assigned to the QA (personally I use it since some weeks)

hum seems I forget to report my test for this one.

so kernel-rt tested this afternoon, Mga2, 64bits, all hardware working fine, dkms are building without issue, latency with qjackctl and lmms seems not that bad.

OK Manuel !
After reading your comment I just saw the update request for it... (bug 8227)

Sorry for the noise !

Nevertheless I will wait kmod is validated before testing the kernels on MGA2 32bits :)

Thanks
Philippe

Or install it and help validate both kmod and kernel-rt :)

Hi Thomas!
I have learned to be cautious ...

I use my everyday computer for tests : so I use to test only packages that may be easily downgraded or suppressed , and that won't need a total re-installation in case of huge problem....

I can't dare to test kmod if there's a risk that my computer won't boot anymore and so there is no way to simply downgrade this package... and a need to reinstall and tune again a perfect Mageia2 
;)

Advisory:
This kernel-rt update provides an upgrade to upstream 3.4-longterm branch.

This allows us to benefit from extra tests and maintenance from upstream
developers and testers, providing a very good base for Mageia users.


It also fixes the following security issues:
- Calling uname() with the UNAME26 personality set allows a leak of
  kernel stack contents. (CVE-2012-0957)

- NFC: Fix multiple remotely-exploitable stack-based buffer overflows due
  to the NCI code pulling length fields directly from incoming frames and
  copying too much data into statically-sized arrays. (CVE-2012-3364)

- A use-after-free flaw has been found in madvise_remove() function in 
  the Linux kernel. madvise_remove() can race with munmap (causing a
  use-after-free of the vma) or with close (causing a use-after-free of
  the struct file). An unprivileged local user can use this flaw to crash
  the system. (CVE-2012-3510)

- Pablo Neira Ayuso discovered that avahi and potentially NetworkManager
  accept spoofed Netlink messages because of a kernel bug. The kernel
  passes all-zero SCM_CREDENTIALS ancillary data to the receiver if the
  sender did not provide such data, instead of not including any such
  data at all or including the correct data from the peer (as it is the
  case with AF_UNIX). (CVE-2012-3520)

- As Tetsuo Handa pointed out, request_module() can stress the system
  while the oom-killed caller sleeps in TASK_UNINTERRUPTIBLE.
  The task T uses "almost all" memory, then it does something which
  triggers request_module().  Say, it can simply call sys_socket().
  This in turn needs more memory and leads to OOM.  oom-killer correctly
  chooses T and kills it, but this can't help because it sleeps in
  TASK_UNINTERRUPTIBLE and after that oom-killer becomes "disabled" by
  the TIF_MEMDIE task T.A local unprivileged user can make the system
  unusable. (CVE-2012-4398)

- A flaw has been found in the way Linux kernel's KVM subsystem handled
  vcpu->arch.cr4 X86_CR4_OSXSAVE bit set upon guest enter. On hosts
  without the XSAVE feature and using qemu userspace an unprivileged
  local user could use this flaw to crash the system. (CVE-2012-4461)

- Commit 644595f89620 ("compat: Handle COMPAT_USE_64BIT_TIME in
  net/socket.c") introduced a bug where the helper functions to take
  either a 64-bit or compat time[spec|val] got the arguments in the
  wrong order, passing the kernel stack pointer off as a user pointer
  (and vice versa).
  Because of the user address range check, that in turn then causes an
  EFAULT due to the user pointer range checking failing for the kernel
  address.  Incorrectly resuling in a failed system call for 32-bit
  processes with a 64-bit kernel.
  On odder architectures like HP-PA (with separate user/kernel address
  spaces), it can be used read kernel memory. (CVE-2012-4467)

- A race condition flaw has been found in the way asynchronous I/O and
  fallocate interacted which can lead to exposure of stale data -- that
  is, an extent which should have had the "uninitialized" bit set
  indicating that its blocks have not yet been written and thus contain
  data from a deleted file. An unprivileged local user could use this
  flaw to cause an information leak. (CVE-2012-4508)

- Reading TCP stats when using TCP Illinois congestion control algorithm
  can cause a divide by zero kernel oops.An unprivileged local user could
  use this flaw to crash the system. (CVE-2012-4565)


Other fixes in this release:
- updates to the upstream 3.4.24-rt36 patchset
- add support for Atheros AR8161/8165 Atheros PCI-E Gigabit Ethernet
  Controller (mga #7853)
- module: fix symbol waiting when module fails before init, wait when
  loading a module which is currently initializing. (mga #7375)

For other changes in 3.4 series kernel, see the kernelnewbies link,
and the referenced stable changelogs.


References:
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-0957
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-3364
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-3510
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-3520
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4398
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4461
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4467
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4508
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4565
http://kernelnewbies.org/Linux_3.4
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.1
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.2
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.3
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.4
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.5
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.6
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.7
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.8
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.9
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.10
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.11
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.12
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.13
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.14
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.15
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.16
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.17
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.18
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.19
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.20
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.21
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.22
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.23
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.24
https://bugs.mageia.org/show_bug.cgi?id=7375
https://bugs.mageia.org/show_bug.cgi?id=7853
https://bugs.mageia.org/show_bug.cgi?id=8068

Doh, last link should be:
https://bugs.mageia.org/show_bug.cgi?id=8515

Additional fixes included:
- backport Wacom Intuos 5 support:
  https://bugs.mageia.org/show_bug.cgi?id=7659

fix for mga #7375 is not in this kernel, will be added later

Updated rpms to validate:

Additional fixes:
- conflict dkms packages not supporting 3.4 series kernels
- kernel-rt-source does not add build & source symlinks anymore as
  they mess with dkms builds...
- A memory disclosure flaw has been found in the way binfmt_script 
  load_script() function handled excessive recursions. An
  unprivileged local user could use this flaw to leak kernel memory.
  (CVE-2012-4530)
- bug 7375 is now fixed here too



i586:
kernel-rt-3.4.24-0.rt36.2.mga2-1-1.mga2.i586.rpm
kernel-rt-devel-3.4.24-0.rt36.2.mga2-1-1.mga2.i586.rpm
kernel-rt-devel-latest-3.4.24-0.rt36.2.mga2.i586.rpm
kernel-rt-doc-3.4.24-0.rt36.2.mga2.noarch.rpm
kernel-rt-latest-3.4.24-0.rt36.2.mga2.i586.rpm
kernel-rt-source-3.4.24-0.rt36.2.mga2-1-1.mga2.noarch.rpm
kernel-rt-source-latest-3.4.24-0.rt36.2.mga2.noarch.rpm



x86_64:
kernel-rt-3.4.24-0.rt36.2.mga2-1-1.mga2.x86_64.rpm
kernel-rt-devel-3.4.24-0.rt36.2.mga2-1-1.mga2.x86_64.rpm
kernel-rt-devel-latest-3.4.24-0.rt36.2.mga2.x86_64.rpm
kernel-rt-doc-3.4.24-0.rt36.2.mga2.noarch.rpm
kernel-rt-latest-3.4.24-0.rt36.2.mga2.x86_64.rpm
kernel-rt-source-3.4.24-0.rt36.2.mga2-1-1.mga2.noarch.rpm
kernel-rt-source-latest-3.4.24-0.rt36.2.mga2.noarch.rpm



SRPMS:
kernel-rt-3.4.24-0.rt36.2.mga2.src.rpm

Summary: Update request: kernel-rt-3.4.24-0.rt36.1.mga2 => Update request: kernel-rt-3.4.24-0.rt36.2.mga2

Updated advisory:

This kernel-rt update provides an upgrade to upstream 3.4-longterm branch.

This allows us to benefit from extra tests and maintenance from upstream
developers and testers, providing a very good base for Mageia users.


It also fixes the following security issues:
- Calling uname() with the UNAME26 personality set allows a leak of
  kernel stack contents. (CVE-2012-0957)

- NFC: Fix multiple remotely-exploitable stack-based buffer overflows due
  to the NCI code pulling length fields directly from incoming frames and
  copying too much data into statically-sized arrays. (CVE-2012-3364)

- A use-after-free flaw has been found in madvise_remove() function in 
  the Linux kernel. madvise_remove() can race with munmap (causing a
  use-after-free of the vma) or with close (causing a use-after-free of
  the struct file). An unprivileged local user can use this flaw to crash
  the system. (CVE-2012-3510)

- Pablo Neira Ayuso discovered that avahi and potentially NetworkManager
  accept spoofed Netlink messages because of a kernel bug. The kernel
  passes all-zero SCM_CREDENTIALS ancillary data to the receiver if the
  sender did not provide such data, instead of not including any such
  data at all or including the correct data from the peer (as it is the
  case with AF_UNIX). (CVE-2012-3520)

- As Tetsuo Handa pointed out, request_module() can stress the system
  while the oom-killed caller sleeps in TASK_UNINTERRUPTIBLE.
  The task T uses "almost all" memory, then it does something which
  triggers request_module().  Say, it can simply call sys_socket().
  This in turn needs more memory and leads to OOM.  oom-killer correctly
  chooses T and kills it, but this can't help because it sleeps in
  TASK_UNINTERRUPTIBLE and after that oom-killer becomes "disabled" by
  the TIF_MEMDIE task T.A local unprivileged user can make the system
  unusable. (CVE-2012-4398)

- A flaw has been found in the way Linux kernel's KVM subsystem handled
  vcpu->arch.cr4 X86_CR4_OSXSAVE bit set upon guest enter. On hosts
  without the XSAVE feature and using qemu userspace an unprivileged
  local user could use this flaw to crash the system. (CVE-2012-4461)

- Commit 644595f89620 ("compat: Handle COMPAT_USE_64BIT_TIME in
  net/socket.c") introduced a bug where the helper functions to take
  either a 64-bit or compat time[spec|val] got the arguments in the
  wrong order, passing the kernel stack pointer off as a user pointer
  (and vice versa).
  Because of the user address range check, that in turn then causes an
  EFAULT due to the user pointer range checking failing for the kernel
  address.  Incorrectly resuling in a failed system call for 32-bit
  processes with a 64-bit kernel.
  On odder architectures like HP-PA (with separate user/kernel address
  spaces), it can be used read kernel memory. (CVE-2012-4467)

- A race condition flaw has been found in the way asynchronous I/O and
  fallocate interacted which can lead to exposure of stale data -- that
  is, an extent which should have had the "uninitialized" bit set
  indicating that its blocks have not yet been written and thus contain
  data from a deleted file. An unprivileged local user could use this
  flaw to cause an information leak. (CVE-2012-4508)

- A memory disclosure flaw has been found in the way binfmt_script 
  load_script() function handled excessive recursions. An
  unprivileged local user could use this flaw to leak kernel memory.
  (CVE-2012-4530)

- Reading TCP stats when using TCP Illinois congestion control algorithm
  can cause a divide by zero kernel oops.An unprivileged local user could
  use this flaw to crash the system. (CVE-2012-4565)


Other fixes in this release:
- module: fix symbol waiting when module fails before init, wait when
  loading a module which is currently initializing. (mga #7375)
- backport Wacom Intuos 5 support: (#7659)
- add support for Atheros AR8161/8165 Atheros PCI-E Gigabit Ethernet
  Controller (mga #7853)
- conflict dkms packages not supporting 3.4 series kernels
- kernel-rt-source does not add build & source symlinks anymore as
  they mess with dkms builds.


For other changes in 3.4 series kernel, see the kernelnewbies link,
and the referenced stable changelogs.


References:
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-0957
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-3364
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-3510
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-3520
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4398
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4461
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4467
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4508
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4530
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-4565
http://kernelnewbies.org/Linux_3.4
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.1
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.2
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.3
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.4
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.5
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.6
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.7
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.8
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.9
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.10
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.11
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.12
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.13
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.14
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.15
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.16
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.17
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.18
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.19
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.20
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.21
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.22
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.23
http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.4.24
https://bugs.mageia.org/show_bug.cgi?id=7375
https://bugs.mageia.org/show_bug.cgi?id=7659
https://bugs.mageia.org/show_bug.cgi?id=7853
https://bugs.mageia.org/show_bug.cgi?id=8068

Created attachment 3364 [details]
dmesg for working kernel-server-3.4.24-3

MGA2 32bits

ASUS M2N SLI mainboard
Nvidia Geforce 210 graphic card
two internal SATA harddisks    (using sata_nv module)
one internal IDE harddisk      (using pata_amd module)
one external usb harddisk 
one external firewire harddisk
one internal IDE DVD writer    (using pata_amd module)
one internal IDE DVD reader    (using pata_amd module)

Nota Bene: I already tested kernel-server-3.4.24-3.mga2-1-1 which is OK ...


testing now :
kernel-rt-3.4.24-0.rt36.2.mga2-1-1.mga2.i586.rpm
kernel-rt-devel-3.4.24-0.rt36.2.mga2-1-1.mga2.i586.rpm

during install the nvidia module is built, using
dkms-nvidia-current-295.71-1.mga2.nonfree.i586 


Boot begins well (same as  kernel-server-3.4.24-3.mga2-1-1.mga2.i586)

But it freezes half the way when starting X :
black screen
no possibility to use Alt+Ctl+Del 
need to Alt+sysRq+r
        Alt+sysRq+s
        Alt+sysRq+e
        Alt+sysRq+i
        Alt+sysRq+u
        Alt+sysRq+b

what would be useful for you besides dmesg files ?

here are attached dmesg for working kernel-server
and dmesg.old for freezing kernel-rt

Created attachment 3365 [details]
dmesg.old for freezing kernel-rt-3.4.24-0rt36.2

Philippe, can you check to see if the freeze is also there
with the prior version of the rt kernel?

The rt kernel worked ok on my system, both x86-64 and i586.

As the kernel update includes security fixes, it will only
be blocked if the freeze is a regression.

CC: (none) => davidwhodgins

Hi David

I had the same freeze problem with kernel-3.2.16-0-rt27.1 on MGA2 32bits...
needing to use magic keys, too, to reboot.

I thought that it was a only a version problem (3.2 serie ) because :
the "normal" kernels version 3.3.6 were OK
and then updated kernels 3.3.8 were OK too...
(in Mandriva 2010.1 "normal" kernels and kernel-rt used to be from the same version and all of them worked for me)

That's the reason why I never wrote a bug report, only hoping that an update to 3.3.x version will appear for kernel-rt... but there had not been any update of the kernel-rt before this one.

So, this freeze is not a regression for me :

But this new version doesn't bring the hoped solution to the problem I had with previous version 3.2.16...   :-(


You may validate it for the security fixes if it's OK for you ...
(I don't think there is such a lot of people needing a true "real-time-kernel" for Mageia2, since "normal" kernels are quite OK for Computer Assisted Music : I have quite no XRuns with the last 3.3.x versions... when there were indeed the need of a real-time-kernel for Mandriva)


Perhaps it's better that I create a new bug report about this freeze for both 
kernel-3.2.16-0-rt27.1.mga2-1-1.mga2
and kernel-rt-3.4.24-0.rt36.2.mga2-1-1.mga2
if Thomas wants and needs to investigate ?! but I certainly will be alone with this bug (probably nvidia module not working for me)
(Nota Bene : it is the same computer, and the dkms nvidia module was OK for both "normal" and rt kernels with Mandriva 2010.1)


Post Scriptum :
Nevertheless, I am very surprised by the huge difference of weight between rt-kernel 2.4Mo   and   server-kernel 3.2Mo !
( in Mandriva the difference was indeed less important, and inverse : 
rt-kernel 2.6Mo   and   desktop-kernel 2.2Mo)

Regards 
Philippe

-rt kernel is xz compressed, -server kernel is gzip compressed.

Ah OK!

Trying to understand where could be the cause of the non working rt-kernel I "kompared" the config files of -rt kernel and server-kernel : 

I already noticed this detail :
68 #CONFIG_KERNEL_GZIP is not set
./.
71 CONFIG_KERNEL_XZ=y

69 CONFIG_KERNEL_GZIP=y
./.
72 #CONFIG_KERNEL_XZ is not set

but I didn't think that would imply such a difference of weight !



Amongst the other differences, I didn't understand which one could explain the freeze   :-(

So, the previous 3.3.8 -server kernel was xz compressed too ...
(same option as 3.3.24 -rt : I just Kompared their config  now !)

Yep. the 3.4.24 -server kernel just about to be pushed has the info:

- switch server kernel image compression back to GZIP so it works
  with Amazon EC2 and other older XEN setups.


so its intentional :)

Thanks Thomas !

As it's not a regression, I'll go ahead and validate the update then.

Could someone from the sysadmin team push the srpm
kernel-rt-3.4.24-0.rt36.2.mga2.src.rpm
from Mageia 2 Core Updates Testing to Core Updates.

See Comment 11 for the advisory.

Keywords: (none) => validated_update
CC: (none) => sysadmin-bugs
Whiteboard: (none) => MGA2-64-OK MGA2-32-OK

Update pushed.

https://wiki.mageia.org/en/Support/Advisories/MGASA-2013-0016

Status: ASSIGNED => RESOLVED
Resolution: (none) => FIXED