Re: [RFC] Linux 2.6.9.1-pre1 contents

Re: [RFC] Linux 2.6.9.1-pre1 contents

[Alan Cox]

On Gwe, 2004-10-29 at 16:04, Chuck Ebbert wrote:
> > This is definitely wrong,
>   It just went into mainline as rev 1.87 to ide-probe.c  :(

Yes its still being discussed and revised. 

>   Your "Integrated Technology Express" addition is missing but I'm
> going to add it here.

You only need that if you have IT8212 support merged (that is in -ac5
again and activated now). It requires the taskfile and ide-disk geometry
changes but does not require all the ide locking stuff (the user
probably does for other reasons but the ide code doesnt)

Alan

Re: [RFC] Linux 2.6.9.1-pre1 contents

[Chuck Ebbert]

On Fri, 29 Oct 2004 at 12:33:04 +0100 Alan Cox wrote:

>> ide_maxtor_probe.patch
>>         Another Maxtor IDE drive serial number oddity
>
> This is definitely wrong,

  It just went into mainline as rev 1.87 to ide-probe.c  :(

  Your "Integrated Technology Express" addition is missing but I'm
going to add it here.


> the others look fairly sane although there are
> a lot that are for such obscure cases they seem to add not reduce risk.

  I've been including one-liners even if they only fix rare cases or
aren't strictly bugfixes, but the larger ones are riskier:

    - backport of the rq-lock patch almost certainly created a profiling
      bug if the retry path is taken.

    - the timers-and-signals patch fixes three things; only one is urgent.

  And Hugh's VM accounting patches look very worthwhile (they just went into
mainline.)  I took one and might merge the other two - but they change
statm reporting.


  Getting them to merge and compile is the easy part...


--Chuck Ebbert  29-Oct-04  11:02:40

Re: [RFC] Linux 2.6.9.1-pre1 contents

[Chuck Ebbert]

On Fri, 29 Oct 2004 at 00:46:32 -0700 Chris Wedgwood wrote:

> On Thu, Oct 28, 2004 at 10:02:58PM -0400, Chuck Ebbert wrote:
>
>> I have been working on a set of patches that I currently call Linux
>> 2.6.9.1.
>
> check -mm kernels or http://linux.bkbits.net/linux-2.5/


  You mean like this?


[me@d4 linux-2.6]$ bk pull -R
Pull http://linux.bkbits.net:8080/linux-2.6
  -> file://home/me/bk-repo/linux-2.6
Nothing to pull.


[me@d4 linux-2.6]$ bk prs -r+ drivers/cdrom/cdrom.c
======== drivers/cdrom/cdrom.c 1.78 ========
D 1.78 04/10/28 00:39:59-07:00 petero2@telia.com[torvalds] 84 83 2/0/3397
P drivers/cdrom/cdrom.c
C Fix incorrect Mt Rainier detection
------------------------------------------------


  I merged the identical change to cdrom.c to my changeset on 04/10/27
at 21:20 -0500, i.e. before Linus did.

  However, I only merge bugfixes, not new features, so LVM still works
in my tree.  :)


--Chuck Ebbert  29-Oct-04  09:27:11

Re: [RFC] Linux 2.6.9.1-pre1 contents

[Alan Cox]

On Gwe, 2004-10-29 at 03:02, Chuck Ebbert wrote:
> 3c59x_pci_disable_device.patch
>         Add missing pci_disable_device() in failure paths
>         NOTE: 3c59x_remove_one() needs disable call, too
>               ...as do tulip, natsemi, etc.

The mm tree with this in has shown some problems which aren't yet
fully explained and are more likely the eeprom thing.

>         Fixes from 2.6.9-ac4

I'll post -ac5 shortly although you seem to have most of that
and a lot more.

> ide_maxtor_probe.patch
>         Another Maxtor IDE drive serial number oddity

This is definitely wrong, the others look fairly sane although there are
a lot that are for such obscure cases they seem to add not reduce risk.

Alan

Re: [RFC] Linux 2.6.9.1-pre1 contents

[Chris Wedgwood]

On Thu, Oct 28, 2004 at 10:02:58PM -0400, Chuck Ebbert wrote:

> I have been working on a set of patches that I currently call Linux
> 2.6.9.1.

check -mm kernels or http://linux.bkbits.net/linux-2.5/

[RFC] Linux 2.6.9.1-pre1 contents

[Chuck Ebbert]

I have been working on a set of patches that I currently call
Linux 2.6.9.1.  What follows is the file 000-MANIFEST, the table
of contents for the patchset.  Since the patchset is currently
65 separate files and the administrator of kernel.org says I
have to be in the MAINTAINERS file to post patches there, I
propose to upload the whole patchset as a uuencoded .tar.bz2
to the mailing list (it is about 42K.)

Comments and suggestions for addition or removal of patches welcomed.

-------------------------------------------------------------------------------

000-MANIFEST

        (this file) Contents of patchset 2.6.9.1-pre1


--- NOTE ----------------------------------
Always apply patches in alphabetical order:
        $ ls <dir>/*.patch | sort | xargs --verbose -n 1 patch -p1 -i 
To un-apply:
        $ ls <dir>/*.patch | sort -r | xargs --verbose -n 1 patch -R -p1 -i 
--- NOTE ----------------------------------


000-Makefile.1-pre1.patch

        Update Makefile version

3c59x_pci_disable_device.patch

        Add missing pci_disable_device() in failure paths
        NOTE: 3c59x_remove_one() needs disable call, too
              ...as do tulip, natsemi, etc.

8250_pnp.patch

        Trivial 8250 Pnp fix

ac3_compiler.h_assembly.patch
ac3_config_via_velocity.patch
ac3_cpia_fixes.patch
ac3_i8042.patch
ac3_i8xx_tco_timer.patch
ac3_moxa_wakeup.patch
ac3_ppp_hangup.patch
ac3_skbuff_tso.patch
ac3_smbfs_request.patch
ac3_vm_io.patch

        Fixes from 2.6.9-ac3
                - "blink keyboard lights on panic" code
                  was removed from the i8042 patch

ac4_aic7xxx.patch
ac4_visor.patch

        Fixes from 2.6.9-ac4

cdrom_mrw.patch

        Some drives were incorrectly identified
          as Mt. Ranier capable

compat_ioctl_tiocsbrk.patch

        Add 32-bit compatible IOCTLS

cross_compile_makefile.patch

        Fix Makefile for some cross-compile configs

decnet_connrefused.patch

        Trivial decnet fix

delay_rq_lock.patch

        Prevent auditing deadlock

dm_duplicate_kfree.patch

        Remove duplicate kfree from dm_target.c

exec_no_binfmt.patch

        Return an error on exec with no binfmt handlers

exec_timers_and_signals.patch

        Fix three bugs in exec with threaded apps

fbdev-fix-logo-vga16fb.patch

        Restore missing penguin logo on vga16fb at boot time
          (included because user complained loudly)

hvsi_oops.patch

        Prevent oops in hvsi driver

i2c_amd_kconfig.patch

        Add missing dependency and clarify comments

ide_maxtor_probe.patch

        Another Maxtor IDE drive serial number oddity

ide_no_chs.patch

        Allow ide drives with no C/H/S info to work

ide_pdc202xx_lba48.patch

        LBA48 fix for PDC20267

ide_smart_threshold.patch

        Fix obvious bug in smart_threshold function

init_poison.patch

        Poison __init memory before freeing

initpipefs_proper.patch

        Make init of pipefs an fs_initcall

ioapic_init_section.patch

        Remove __init from referenced ioapic functions

ioapic_on_nvidia_boards.patch

        Ignore all ACPI timer overrides on nVIDIA x86-64 boards
        (Note: enable of IO-APIC in original patch removed
         due to bug reports)

log_buf_shift.patch

        Allow larger kernel log buffer

media_toaster_share_irq.patch

        Permit IRQ sharing on Media Toaster (sym53c500_cs)

memset_signal_ppc64.patch

        Correct memset size in ppc64 signal code

msleep_at_least_amount.patch

        make msleep sleep at least the amount of time asked for

negative_statm.patch

        Don't allow statm fields to go negative

netif_rx_ni_preempt_safe.patch

        Make netif_rx_ni preempt-safe
        NOTE: A better (uninlined) fix is in 2.6.10-rc

netpoll_setup_crash.patch

        Prevent crash on netpoll setup

nfs_fqdn_len_2.patch

        Allow longer FQDN in NFS mount

o_direct_mmapped_io.patch

        Fix O_DIRECT vs. mmapped IO problem
        (Not a complete fix, but at least an improvement)

parport_superio.patch

        Gets superio parports working again

pcdp_swap_args.patch

        Correct argument order in pcdp.c

pci_dev_put.patch

        Define pci_dev_put() for non-PCI builds

percpu_alignment.patch

        Make percpu alignment work right on x86 cpus

pnpbios_msleep.patch

        Use msleep_interruptible instead of schedule_timeout

proc_wrong_parent.patch

        Backport BK changesets 1.2000.4.120 and 1.2000.12.16
        Fixes wrong parent process display in /proc

quoted_env_vars_3.patch

        Unquote kernel arg strings

return_enfile_not_emfile.patch

        return -ENFILE when out of handles, not -EMFILE

s390_sacf_exception.patch

        Fix s390 security hole

spurious_oomkill.patch

        Stop spurious out-of-memory process kills

suspend_time_adjust.patch

        Correct system time after suspend

tcp_output_skbuff_fix.patch

        With TSO, skbuffs were allocated with wrong size

tso_needs_sg.patch

        TSO feature requires SG

unbalanced_tasks_v3.patch

        Scheduler fix for pinned tasks causing too much rebalance

usr_courier_pnp_id.patch

        Add USR Courier PnP modem ID

vga_console_font.patch

        Fix two small bugs in VGA console
        
vm_dirty_ratio_clamp.patch

        Prevent divide-by-zero

vm_pages_scanned_active_list.patch

        kswapd: pages scanned were not counted properly

x86_64_syscall32_initdata.patch

        More __init/__initdata fixes



--Chuck Ebbert  28-Oct-04  20:16:13