]> git.neil.brown.name Git - history.git/log
history.git
23 years agoia64: Checkin support files for vendor-specific ACPI extensions.
David Mosberger [Thu, 3 Apr 2003 04:32:42 +0000 (20:32 -0800)]
ia64: Checkin support files for vendor-specific ACPI extensions.
      Fix spelling of Hewlett-Packard.

23 years ago[PATCH] ia64: fix missing symbol exports
Xavier Bru [Thu, 3 Apr 2003 04:19:17 +0000 (20:19 -0800)]
[PATCH] ia64: fix missing symbol exports

3) with CONFIG_NUMA set, there are undefined symbols building modules:

scripts/modpost vmlinux drivers/md/dm-mod.o drivers/net/e1000/e1000.o drivers/net/eepro100.o drivers/md/md.o drivers/net/mii.o fs/xfs/xfs.o
*** Warning: cpu_info__per_cpu [fs/xfs/xfs.ko] undefined!
*** Warning: cpu_info__per_cpu [drivers/md/md.ko] undefined!
*** Warning: cpu_to_node_map [drivers/md/md.ko] undefined!

exporting them in ia64_ksyms.c fixes the problem.

23 years agoia64: Fix IA64_FETCHADD() macro.
David Mosberger [Mon, 31 Mar 2003 08:45:01 +0000 (00:45 -0800)]
ia64: Fix IA64_FETCHADD() macro.

23 years ago[PATCH] ia64: replace cpu_is_online with cpu_online
Martin Hicks [Mon, 31 Mar 2003 07:58:16 +0000 (23:58 -0800)]
[PATCH] ia64: replace cpu_is_online with cpu_online

Consolidate the uses of
cpu_is_online() to use the (relatively) new
cpu_online() macro.

23 years ago[PATCH] ia64: machine vectors for readX routines
Jesse Barnes [Mon, 31 Mar 2003 07:45:09 +0000 (23:45 -0800)]
[PATCH] ia64: machine vectors for readX routines

We need readX() to be machine vectors since some platforms don't provide
DMA coherence via PIO reads (PCI drivers and the spec imply that this is
a good idea).  Writes are ok though for all existing ia64 platforms (and
hopefully it'll stay that way).

23 years agoia64: Change struct ia64_fpreg so it will get 16-byte alignment with all
David Mosberger [Thu, 27 Mar 2003 09:03:02 +0000 (01:03 -0800)]
ia64: Change struct ia64_fpreg so it will get 16-byte alignment with all
ia64 compilers, not just GCC.

23 years agoia64: Patch by Andreas Schwab to fix sys32_ptrace().
David Mosberger [Thu, 27 Mar 2003 04:49:33 +0000 (20:49 -0800)]
ia64: Patch by Andreas Schwab to fix sys32_ptrace().

23 years ago[PATCH] ia64: Fix request_module from ia32 process
Andreas Schwab [Thu, 27 Mar 2003 04:43:01 +0000 (20:43 -0800)]
[PATCH] ia64: Fix request_module from ia32 process

When an ia32 process triggers request_module the kernel cannot execute
modprobe because the kernel thread still has the ia32 address limits in
force.  I think a kernel thread should always have the ia64 address
limits, similar to what sys32_execve is doing.

23 years ago[PATCH] ia64: don't try to synchronize ITCs on ITC_DRIFT platforms
Jes Sorensen [Thu, 27 Mar 2003 04:38:22 +0000 (20:38 -0800)]
[PATCH] ia64: don't try to synchronize ITCs on ITC_DRIFT platforms

A small patch for 2.4 that stops the kernel from trying to syncrhonize
ITC clocks between CPUs if we know that the ITC isn't synchronized
across the backplane.

23 years ago[PATCH] ia64: update email address
Kochi Takayoshi [Thu, 27 Mar 2003 04:36:27 +0000 (20:36 -0800)]
[PATCH] ia64: update email address

My e-mail address changed and I'd like to update it in arch/ia64/kernel/acpi.c.

23 years agoia64: Two small MCA fixes.
David Mosberger [Thu, 27 Mar 2003 04:34:49 +0000 (20:34 -0800)]
ia64: Two small MCA fixes.

23 years ago[PATCH] ia64: fix unwinder bug in unw_access_gr()
Andreas Schwab [Wed, 26 Mar 2003 06:22:31 +0000 (22:22 -0800)]
[PATCH] ia64: fix unwinder bug in unw_access_gr()

I've found this by inspecting the code: pt_regs_off returns an offset into
struct pt_regs.  Relative to latest bk, 2.5 has the same bug.

23 years ago[PATCH] ia64: fix settimeofday() not synchronised with gettimeofday()
Eric Piel [Wed, 26 Mar 2003 06:18:34 +0000 (22:18 -0800)]
[PATCH] ia64: fix settimeofday() not synchronised with gettimeofday()

Eric Piel wrote:
> However, now, it still gives negative difference:
> # ./a.out
> requested:      1047572128s 2564ns
> new:            1047572128s 1588ns
> diff is -0.000976000sec
>
> That's better but there is still something...
> Can anyone reproduce this bug? Any idea about what may cause this
> shifted results?
>
> I don't understand what does the line in settimeofday():
>         nsec -= (jiffies - wall_jiffies ) * (1000000000 / HZ);

Finally I read the code to do the same thing for i386 (get and
settimeofday()). This explains the meaning of this line, in the i386
it's associated with the equivalent line in do_gettimeofday()! On ia64
everything is done inside of gettimeoffset(). Therefore I'm now
confident that suppressing this line is a Good Thing ;-) The patch doing
it wrt the bk tree is attached.

The test case confirms that it works:
requested:      1048681051s 194873ns
new:            1048681051s 194874ns
diff is  0.000001000sec

That's the same result than on a 2.4.19 .
This also solved an error on the high resolution timers test suite.

23 years agoia64: More module-loader fixing.
David Mosberger [Wed, 26 Mar 2003 04:11:35 +0000 (20:11 -0800)]
ia64: More module-loader fixing.

23 years agoia64: Fix module loader by setting sh_type of place-holder sesctions to SHT_NOBITS.
David Mosberger [Wed, 26 Mar 2003 03:36:28 +0000 (19:36 -0800)]
ia64: Fix module loader by setting sh_type of place-holder sesctions to SHT_NOBITS.

23 years agoia64: Minor fixes.
David Mosberger [Wed, 26 Mar 2003 02:21:11 +0000 (18:21 -0800)]
ia64: Minor fixes.

23 years agoia64: Add module.lds.
David Mosberger [Wed, 26 Mar 2003 01:31:03 +0000 (17:31 -0800)]
ia64: Add module.lds.

23 years agoia64: Fix typo in sys_clone().
David Mosberger [Wed, 26 Mar 2003 01:14:41 +0000 (17:14 -0800)]
ia64: Fix typo in sys_clone().

23 years agoia64: Add ia64-specific LDFLAGS_MODULE and export unwind API to modules.
David Mosberger [Wed, 26 Mar 2003 00:30:49 +0000 (16:30 -0800)]
ia64: Add ia64-specific LDFLAGS_MODULE and export unwind API to modules.

23 years agoia64: Rewrite the relocator in the kernel module loader. Fix some bugs and simplify
David Mosberger [Tue, 25 Mar 2003 17:08:09 +0000 (09:08 -0800)]
ia64: Rewrite the relocator in the kernel module loader.  Fix some bugs and simplify
the handling of loader-created sections.

23 years ago[PATCH] ia64: Cross-compile fix
Wolfgang Mauerer [Mon, 24 Mar 2003 05:56:17 +0000 (21:56 -0800)]
[PATCH] ia64: Cross-compile fix

the attached patch (against bk-current) fixes a cross
compilation problem by using the target specific objdump
tool instead of the host specific one.

23 years agoia64: Manual merge of Keith Owen's patch to avoid deadlock on
David Mosberger [Mon, 24 Mar 2003 05:43:45 +0000 (21:43 -0800)]
ia64: Manual merge of Keith Owen's patch to avoid deadlock on
ia64_sal_mc_rendez().  Also prefix local-variables in
SAL macros to avoid name collisions.

23 years ago[PATCH] ia64: mca rendezvous fix
Keith Owens [Mon, 24 Mar 2003 05:36:20 +0000 (21:36 -0800)]
[PATCH] ia64: mca rendezvous fix

We are not setting the 'always rendezvous for mca' flag.  kdb needs it
set to get decent mca debugging on all cpus but I do not want kdb to
change sal behaviour.  Since we do not recover from mca without a
debugger, I see no reason why this flag should not be on for all
kernels.

The rendezvous timeout was set to 100 * HZ, but SAL expects the timeout
to be in milliseconds, HZ may not be 1 millisecond.  The patch makes
the timeout an explicit 20 seconds, semi-arbitrary value.

23 years agoFirst draft at making modules work again (loosely based on Rusty's original and
David Mosberger [Fri, 21 Mar 2003 18:53:01 +0000 (10:53 -0800)]
First draft at making modules work again (loosely based on Rusty's original and
thoroughly broken ia64 patch).  Not all relocs are supported yet and the reloc
code needs to be cleaned up, but simply stuff like loading the palinfo module
works.  Also, linkage-stubs are optimized with brl for McKinley or better.

23 years agoia64: Minor Makefile cleanup.
David Mosberger [Fri, 21 Mar 2003 18:46:33 +0000 (10:46 -0800)]
ia64: Minor Makefile cleanup.

23 years ago[PATCH] ia64: improve show_trace_task() portability
Vladimir Simonov [Fri, 21 Mar 2003 03:44:54 +0000 (19:44 -0800)]
[PATCH] ia64: improve show_trace_task() portability

Trying to port some patches from i386 to ia64 I've found that
show_trace_task is the only portable way to show tasks' call
traces(perhaps because it is called from ./kernel/sched.c:)).
But its ia64 implementation can't work with running task.

Attached patch fixes this issue.

As I understand show_trace and show_stack are platform
dependent or have different sense/args.

23 years ago[PATCH] ia64: Use PAL_HALT_LIGHT in cpu_idle
Alex Williamson [Wed, 19 Mar 2003 01:24:26 +0000 (17:24 -0800)]
[PATCH] ia64: Use PAL_HALT_LIGHT in cpu_idle

Here's patches for 2.4 & 2.5 to use PAL_HALT_LIGHT in cpu_idle.
This helps to reduce CPU temp a little on boxes with firmware that
takes advantage of this lower power state.  I've tried this on a
rx2600 (2x900MHz McKinley) and an i2000 (fw 117) and it shows some
benefit.  On McKinley systems, only the very latest PAL from Intel
actually reduces power consumption in the halt_light state.  For HP
rx2600/zx6000/zx2000, this means you need to be running firmware 1.82.

  Rohit Seth, at Intel, has run some benchmarks with this kind of
modification and found the effects of enabling halt_light to fall
within the noise of mosts tests.  I replaced pal_halt(1) in safe_halt
with pal_halt_light() since halt_light is required to be implemented,
but pal_halt(1) is an optional halt state.  I'd be interested to hear
of any measurements anyone does using this, where it works/fails, and
if any benchmarks/applications are impacted.

23 years ago[PATCH] ia64: SN makefile update take
Jesse Barnes [Tue, 18 Mar 2003 14:45:06 +0000 (06:45 -0800)]
[PATCH] ia64: SN makefile update take

David, here's an updated patch that doesn't mess with
arch/ia64/Makefile in case you hadn't applied the last one yet

23 years ago[PATCH] ia64: print ISR for FPSWA faults
Martin Hicks [Tue, 18 Mar 2003 14:22:40 +0000 (06:22 -0800)]
[PATCH] ia64: print ISR for FPSWA faults

Here is a simple patch to also print isr during the handling of a
floating point assist fault.

23 years ago[PATCH] ia64: remove stale mmiob function
Jesse Barnes [Tue, 18 Mar 2003 07:12:07 +0000 (23:12 -0800)]
[PATCH] ia64: remove stale mmiob function

The consensus on lkml was that devices should do reads from safe
registers to ensure PIO write ordering, which means we no longer need
mmiob.  This patch removes the mmiob entries from the machine vector
headers and io.h and updates the documentation about PIO ordering.

23 years agoia64: Patch by Andreas Schwab: The read_lock and read_unlock macros
David Mosberger [Tue, 18 Mar 2003 06:59:30 +0000 (22:59 -0800)]
ia64: Patch by Andreas Schwab: The read_lock and read_unlock macros
should not use such innocent variable names like tmp because they
have a high probability to clash with (part of) the argument.

23 years ago[PATCH] ia64: unwind.c - allow unw_access_gr(r0)
Keith Owens [Fri, 14 Mar 2003 03:25:24 +0000 (19:25 -0800)]
[PATCH] ia64: unwind.c - allow unw_access_gr(r0)

The patch allows unw_access_gr() to read from r0, to support unwind
directives such as .save ar.pfs,r0 and .save rp,r0.

23 years agoia64: Fix sys_clone() to take 5 input arguments.
David Mosberger [Fri, 14 Mar 2003 03:24:47 +0000 (19:24 -0800)]
ia64: Fix sys_clone() to take 5 input arguments.

23 years ago[PATCH] ia64: POSIX timer fixes
Eric Piel [Thu, 13 Mar 2003 05:41:41 +0000 (21:41 -0800)]
[PATCH] ia64: POSIX timer fixes

Here is a patch to have the POSIX timer interface completly integrated
in ia64 (2.5.64). The programs in userland can now access the siginfo
structure. With that patch the test programs of the high resolution
timers pass without error but one which seems to also be triggered on
ix86: nanosleeps too short.

23 years agoia64: Fix settimeofday(). Based on patch by Eric Piel.
David Mosberger [Thu, 13 Mar 2003 04:13:13 +0000 (20:13 -0800)]
ia64: Fix settimeofday().  Based on patch by Eric Piel.

23 years ago[PATCH] ia64: clone2/clone argument order fixes
Jakub Jelínek [Thu, 13 Mar 2003 03:34:20 +0000 (19:34 -0800)]
[PATCH] ia64: clone2/clone argument order fixes

do_fork is declared as:
struct task_struct *do_fork(unsigned long clone_flags,
                            unsigned long stack_start,
                            struct pt_regs *regs,
                            unsigned long stack_size,
                            int *parent_tidptr,
                            int *child_tidptr)
ie. parent_tidptr is out4 and child_tidptr is out5, but
the comments in clone2 were suggesting otherwise.
So, we either need a patch which will codify current order
of clone arguments (ie. ptid, ctid, tls; attached below - has the advantage
that clone2 stays with the same ABI as in 2.5.[56]x), or the arguments
of clone2 and clone should be reordered to match the IA-32 order
(which is ptid, tls, ctid).

23 years agoia64: Rename __put_task_struct() to free_task_struct(). Based on patch by Peter...
David Mosberger [Tue, 11 Mar 2003 15:40:56 +0000 (07:40 -0800)]
ia64: Rename __put_task_struct() to free_task_struct().  Based on patch by Peter Chubb.

23 years agoia64: Remove ia64_spinlock_contention().
David Mosberger [Tue, 11 Mar 2003 09:51:54 +0000 (01:51 -0800)]
ia64: Remove ia64_spinlock_contention().

23 years ago[PATCH] ia64: declare test_bit() arg as "const"
Peter Chubb [Tue, 11 Mar 2003 08:47:38 +0000 (00:47 -0800)]
[PATCH] ia64: declare test_bit() arg as "const"

While you're fixing bitops.h, making test_bit take a const qualified
arg will kill some warnings in reiserfs...

23 years ago[PATCH] ia64: CPE & CMC polling for 2.5
Alex Williamson [Tue, 11 Mar 2003 08:08:19 +0000 (00:08 -0800)]
[PATCH] ia64: CPE & CMC polling for 2.5

Here's another feature I'd like to add to MCA support; the ability
to detect a flood of CMCs and switch to polling mode for retrieving
CMC logs.  Once no more CMC logs are found, return to and interrupt
driven handler.  If the flood threshold is never reached, the CMC
handler simply behaves as it does today.

   It's useful to get the CMC logs to know that something isn't
quite right, but if you end up with some bad memory it's too easy for
them to interfere with useful work.  I've tested this on an HP rx2600,
with a known bad DIMM.  This DIMM acts like it has a completely dead
DRAM on it.  With the current CMC handler, once I hit that range of
memory addresses, the system essentially dies, constantly handling
CMC errors.  With this patch, the system hits the threshold quickly,
but remains functional with no performance degredation once in polling
mode.  This patch applies against linux-2.4.20-ia64-021210 and
includes:

 - Switching CMCs to polling mode at predeterimined threshold
 - If polling for CPEs, poll on all processors
 - Fix timestamp on log output

23 years agoia64: Drop unused NEW_LOCK spinlock code and clean up unneeded test
David Mosberger [Tue, 11 Mar 2003 06:39:33 +0000 (22:39 -0800)]
ia64: Drop unused NEW_LOCK spinlock code and clean up unneeded test
in kernel unwinder.

23 years ago[PATCH] ia64: IA32 support without sysctl doesn't work lia64-v2.5.64
Bjorn Helgaas [Thu, 6 Mar 2003 14:28:14 +0000 (06:28 -0800)]
[PATCH] ia64: IA32 support without sysctl doesn't work

Make CONFIG_SYSCTL control sys32_sysctl as well.  Based on a patch from Peter Chubb.

23 years agoia64: Minor formatting/whitespace fixes in ia64-version of acpi.c.
David Mosberger [Thu, 6 Mar 2003 14:21:44 +0000 (06:21 -0800)]
ia64: Minor formatting/whitespace fixes in ia64-version of acpi.c.

23 years ago[PATCH] ia64: ACPI fix for no PCI
Jesse Barnes [Thu, 6 Mar 2003 14:17:03 +0000 (06:17 -0800)]
[PATCH] ia64: ACPI fix for no PCI

Andy Grover told me this should be posted here.  It allows ACPI to
compile even with PCI turned off.  Patch against 2.5.60.

23 years ago[PATCH] ia64: compat_sys_futex() support
Stephen Rothwell [Thu, 6 Mar 2003 14:04:54 +0000 (06:04 -0800)]
[PATCH] ia64: compat_sys_futex() support

ia 64 part of the patch.  This gives you a 32 bit version of sys_futex (hopefully).

23 years agoia64: Fix several small bugs/omissions from the 2.5.64 sync.
David Mosberger [Thu, 6 Mar 2003 13:11:54 +0000 (05:11 -0800)]
ia64: Fix several small bugs/omissions from the 2.5.64 sync.

23 years agoia64: Hook up POSIX-timer syscalls. Take advantage of ptrace_notify().
David Mosberger [Thu, 6 Mar 2003 06:56:39 +0000 (22:56 -0800)]
ia64: Hook up POSIX-timer syscalls.  Take advantage of ptrace_notify().

23 years agoia64: Correct the value of siginfo.si_addr for SIGSEGV signals
David Mosberger [Thu, 6 Mar 2003 05:42:13 +0000 (21:42 -0800)]
ia64: Correct the value of siginfo.si_addr for SIGSEGV signals
triggered by NaT-page-consumption faults.

23 years agoia64: Sync with 2.5.64.
David Mosberger [Thu, 6 Mar 2003 05:34:32 +0000 (21:34 -0800)]
ia64: Sync with 2.5.64.

23 years agoia64: Make ia64_fetch_and_add() simpler to optimize so lib/rwsem.c lia64-v2.5.60
David Mosberger [Thu, 6 Mar 2003 03:22:26 +0000 (19:22 -0800)]
ia64: Make ia64_fetch_and_add() simpler to optimize so lib/rwsem.c
can be optimized properly.

23 years agoLinux 2.5.64 v2.5.64
Linus Torvalds [Tue, 4 Mar 2003 11:11:19 +0000 (03:11 -0800)]
Linux 2.5.64

23 years agoia64: Implement pcibios_prep_mwi() and define HAVE_ARCH_PCI_MWI to
David Mosberger [Tue, 4 Mar 2003 10:40:21 +0000 (02:40 -0800)]
ia64: Implement pcibios_prep_mwi() and define HAVE_ARCH_PCI_MWI to
ensure that PCI line-size gets programmed properly.  Based
on patch by Grant Grundler.

23 years agoia64: Implement _raw_write_trylock(). Based on patch by Joel Guillet.
David Mosberger [Tue, 4 Mar 2003 10:38:51 +0000 (02:38 -0800)]
ia64: Implement _raw_write_trylock().  Based on patch by Joel Guillet.

23 years agoia64: Fix SAL processor-log info handling. Based on patch by
David Mosberger [Tue, 4 Mar 2003 10:34:01 +0000 (02:34 -0800)]
ia64: Fix SAL processor-log info handling.  Based on patch by
Keith Owens.

23 years agoia64: Fix ia32 sysinfo() emulation.
David Mosberger [Tue, 4 Mar 2003 10:28:10 +0000 (02:28 -0800)]
ia64: Fix ia32 sysinfo() emulation.

23 years agoia64: Minor formatting fixes for the preemption patch.
David Mosberger [Tue, 4 Mar 2003 10:16:04 +0000 (02:16 -0800)]
ia64: Minor formatting fixes for the preemption patch.

23 years ago[PATCH] ia64: Preemption patch against ~2.5.60
Peter Chubb [Tue, 4 Mar 2003 10:06:48 +0000 (02:06 -0800)]
[PATCH] ia64: Preemption patch against ~2.5.60

Latest premption patch.

23 years ago[PATCH] ia64: 2nd update for HugeTLB Page patch for IA-64 2.5.60 kernel
Seth Rohit [Tue, 4 Mar 2003 09:33:01 +0000 (01:33 -0800)]
[PATCH] ia64: 2nd update for HugeTLB Page patch for IA-64 2.5.60 kernel

Please find attached a small hugetlb page support update for 2.5.60
IA-64 kernel (+ my previous hugetlb page patch).  This patch adds the
HAVE_ARCH_HUGETLB_UNMAPPED_AREA define for IA-64 arch.
hugetlb_get_unmapped_area function is added in
arch/ia64/mm/hugetlbpage.c

23 years ago[PATCH] ia64: HugeTLB Page patch for IA-64 2.5.60 kernel
Seth Rohit [Tue, 4 Mar 2003 09:30:54 +0000 (01:30 -0800)]
[PATCH] ia64: HugeTLB Page patch for IA-64 2.5.60 kernel

Please find attached a hugetlb page patch for IA-64 2.5.60 kernel .
Changes in generic files are mostly backported from 2.5.62 (to get
hugetlb support functioning properly).

23 years ago[PATCH] More IEEE1394 updates
Ben Collins [Tue, 4 Mar 2003 08:25:16 +0000 (00:25 -0800)]
[PATCH] More IEEE1394 updates

 - Cleanup Makefile.
 - Fix potentional crash in dv1394 module unload.
 - Why does ARM's memcpy not return void*? Oh well, work around it in
  eth1394 by not expecting standard behavior.

23 years agoFix up sunrpc for the dentry hash list changes.
Linus Torvalds [Tue, 4 Mar 2003 07:46:50 +0000 (23:46 -0800)]
Fix up sunrpc for the dentry hash list changes.

23 years ago[PATCH] ia64: SN update
Jesse Barnes [Tue, 4 Mar 2003 07:00:36 +0000 (23:00 -0800)]
[PATCH] ia64: SN update

And here's the SN specific part of the update.  This should get an SN2
compile all the way to the link stage, where I still have some devfs
stuff to cleanup.

23 years ago[PATCH] ia64: SN updates for mmzone.h
Jesse Barnes [Tue, 4 Mar 2003 06:59:02 +0000 (22:59 -0800)]
[PATCH] ia64: SN updates for mmzone.h

Here's an update to the ia64 mmzone.h file that adds some SN2 specific
defines (which will probably have to be changed later, but it's a
start).

23 years ago[PATCH] ia64: 2/2 fix in machvec.h
Suresh B. Siddha [Tue, 4 Mar 2003 06:55:41 +0000 (22:55 -0800)]
[PATCH] ia64: 2/2 fix in machvec.h

IA64 ABI specifies that globals > 8 bytes need to be aligned to 16 bytes.
gcc doesn't follow this convention. Current kernel code will fail to work with
a compiler which follows the ABI.

size of structure ia64_machine_vector is > 8 bytes and not multiple of 16 bytes.
When we have CONFIG_IA64_GENERIC, each machine specific vector from different
object files gets linked into a user defined section(forming array of
structures).

Now with a compiler conforming to ABI, there will be holes in this array
resulting in the code failure when it goes through this array of structures.

Following patch will make the size of the structure to be multiple of 16 bytes
making both ABI confirming/non-conforming compilers happy.

23 years ago[PATCH] ia64: 1/2 fix for generic kernel
Suresh B. Siddha [Tue, 4 Mar 2003 06:52:56 +0000 (22:52 -0800)]
[PATCH] ia64: 1/2 fix for generic kernel

Current 2.5 kernels are broken with CONFIG_IA64_GENERIC option. Two unused
members are still remaining in the structure, resulting in corruption of
structure elements during the MACHVEC_INIT initialization. Attached patch
fixes this problem.

23 years agoia64: Correct region_start calculation in kernel unwinder.
David Mosberger [Tue, 4 Mar 2003 06:51:00 +0000 (22:51 -0800)]
ia64: Correct region_start calculation in kernel unwinder.

23 years ago[PATCH] ia64: iosapic: rationalize __init/__devinit
Bjorn Helgaas [Tue, 4 Mar 2003 06:44:05 +0000 (22:44 -0800)]
[PATCH] ia64: iosapic: rationalize __init/__devinit

Rationalize __init/__devinit attributes.  The noteworthy changes are
that
iosapic_system_init(),
iosapic_init(),
iosapic_register_platform_intr(), and
iosapic_override_isa_irq()
are __init (only called from ACPI __init functions), but
iosapic_lists[],
num_iosapic,
find_iosapic(),
register_intr(),
iosapic_register_intr(), and
acpi_register_irq()
are not because they may be used after init-time by modules.

More detailed analysis:

        iosapic_lists[], num_iosapic: normal, referenced by
                find_iosapic (normal)
                register_intr (normal)
                iosapic_init (__init)

        pcat_compat: __initdata, referenced by
                iosapic_system_init (__init)
                iosapic_init (__init)
                iosapic_parse_prt (__init)

        find_iosapic: normal, called by
                register_intr (normal)

        register_intr: normal, called by
                iosapic_register_intr (normal), called by
                        acpi_register_irq (normal), called by
                                modules (=> can't be __init or __devinit)
                iosapic_register_platform_intr (__init)
                iosapic_override_isa_irq (__init)
                iosapic_parse_prt (__init)

        iosapic_reassign_vector: __init, called by
                iosapic_register_platform_intr (__init), called by
                        acpi_parse_plat_int_src (__init)

        iosapic_system_init: __init, called by
                acpi_parse_madt (__init)

        iosapic_init: __init, called by
                acpi_parse_iosapic (__init)

        iosapic_register_platform_intr: __init, called by
                acpi_parse_plat_int_src (__init)

        iosapic_override_isa_irq: __init, called by
                acpi_parse_int_src_ovr (__init)
                iosapic_init (__init)

        fixup_vector: __init, called by
                iosapic_parse_prt (__init), called by
                        acpi_pci_irq_init (__init)

23 years ago[PATCH] ia64: iosapic: self-documenting polarity/trigger arguments
Bjorn Helgaas [Tue, 4 Mar 2003 06:41:55 +0000 (22:41 -0800)]
[PATCH] ia64: iosapic: self-documenting polarity/trigger arguments

Make interrupt registration functions take named constants for
polarity and trigger mode.  Old -> new magic decoder ring:
polarity 0 -> IOSAPIC_POL_LOW(#defined to 1)
polarity 1 -> IOSAPIC_POL_HIGH(#defined to 0)
trigger 0  -> IOSAPIC_LEVEL(#defined to 1)
trigger 1  -> IOSAPIC_EDGE(#defined to 0)

23 years ago[PATCH] ia64: iosapic: simplify ISA IRQ init
Bjorn Helgaas [Tue, 4 Mar 2003 06:39:32 +0000 (22:39 -0800)]
[PATCH] ia64: iosapic: simplify ISA IRQ init

Simplify ISA IRQ init by taking advantage of iosapic_override_isa_irq(),
which already does what we need.

23 years ago[PATCH] ia64: iosapic: remove find_iosapic duplication
Bjorn Helgaas [Tue, 4 Mar 2003 06:38:31 +0000 (22:38 -0800)]
[PATCH] ia64: iosapic: remove find_iosapic duplication

emove IOSAPIC address and GSI base from external interrupt
registration interfaces.  This lets us remove acpi_find_iosapic(),
which is functionally similar to find_iosapic().

23 years ago[PATCH] ia64: iosapic: make pcat_compat system property
Bjorn Helgaas [Tue, 4 Mar 2003 06:36:48 +0000 (22:36 -0800)]
[PATCH] ia64: iosapic: make pcat_compat system property

Make pcat_compat a system property, not a per-IOSAPIC property.

23 years agoMerge penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/hlist
Linus Torvalds [Tue, 4 Mar 2003 06:30:19 +0000 (22:30 -0800)]
Merge penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/hlist
into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

23 years ago[PATCH] dcache/inode hlist patchkit
Andi Kleen [Tue, 4 Mar 2003 06:28:32 +0000 (22:28 -0800)]
[PATCH] dcache/inode hlist patchkit

 - Inode and dcache Hash table only needs half the memory/cache because
   of using hlists.
 - Simplify dcache-rcu code.  With NULL end markers in the hlists
   is_bucket is not needed anymore.  Also the list walking code
   generates better code on x86 now because it doesn't need to dedicate
   a register for the list head.
 - Reorganize struct dentry to be more cache friendly.  All the state
   accessed for the hash walk is in one chunk now together with the
   inline name (all at the end)
 - Add prefetching for all the list walks.  Old hash lookup code didn't
   use it.
 - Some other minor cleanup.

23 years agoia64: In kernel unwinder, replace dump_info_pt() with get_scratch_regs()
David Mosberger [Tue, 4 Mar 2003 06:27:36 +0000 (22:27 -0800)]
ia64: In kernel unwinder, replace dump_info_pt() with get_scratch_regs()
and reformat to make it fit in 100 columns.

23 years agoia64; Improve debug output from kernel unwinder. Based on patch by
David Mosberger [Tue, 4 Mar 2003 05:58:54 +0000 (21:58 -0800)]
ia64; Improve debug output from kernel unwinder.  Based on patch by
Keith Owens.

23 years ago[PATCH] ia64: fix scratch-regs handling in kernel unwinder
Keith Owens [Tue, 4 Mar 2003 05:44:06 +0000 (21:44 -0800)]
[PATCH] ia64: fix scratch-regs handling in kernel unwinder

This patch has been running inside SGI for 2 months.  It
handles kernel stacks with multiple struct pt_regs, as found
while debugging the kernel.

23 years ago[PATCH] ia64: rwsem using atomic primitive
Kenneth W. Chen [Tue, 4 Mar 2003 04:05:08 +0000 (20:05 -0800)]
[PATCH] ia64: rwsem using atomic primitive

23 years ago[PATCH] ia64: CONFIG_NUMA build fix
Junichi Nomura [Tue, 4 Mar 2003 03:59:32 +0000 (19:59 -0800)]
[PATCH] ia64: CONFIG_NUMA build fix

I had to apply the patch below to build with CONFIG_NUMA.

23 years agoMerge bk://ldm.bkbits.net/linux-2.5-core
Linus Torvalds [Tue, 4 Mar 2003 03:09:35 +0000 (19:09 -0800)]
Merge bk://ldm.bkbits.net/linux-2.5-core
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoMerge osdl.org:/home/mochel/src/kernel/devel/linux-2.5-virgin
Patrick Mochel [Tue, 4 Mar 2003 06:17:16 +0000 (00:17 -0600)]
Merge osdl.org:/home/mochel/src/kernel/devel/linux-2.5-virgin
into osdl.org:/home/mochel/src/kernel/devel/linux-2.5-core

23 years agosysfs: Register filesystems with sysfs.
Patrick Mochel [Tue, 4 Mar 2003 06:13:14 +0000 (00:13 -0600)]
sysfs: Register filesystems with sysfs.

Originally from Nikita Danliov.

- Add struct subsystem to struct file_system_type.
- Add fs_subsys_init() to explicitly call sysfs_init() very early.
- Register base fs subsystem then, too.
- Register filesystems' embedded subsystem when registered with fs core.

Resulting tree looks like:

# tree /sys/fs/
/sys/fs/
|-- bdev
|-- devpts
|-- eventpollfs
|-- ext2
|-- ext3
|-- futexfs
|-- iso9660
|-- pipefs
|-- proc
|-- ramfs
|-- rootfs
|-- sockfs
|-- sysfs
`-- tmpfs

23 years agodriver model: implement platform_match()
Patrick Mochel [Tue, 4 Mar 2003 04:20:36 +0000 (22:20 -0600)]
driver model: implement platform_match()

- For matching registered platform devices with platform drivers.

- Also, make sure platform_bus is initialized early, with the rest of the
  core stuff.

23 years agosysfs: initialize from fs/namespace.c::mnt_init().
Patrick Mochel [Tue, 4 Mar 2003 04:17:19 +0000 (22:17 -0600)]
sysfs: initialize from fs/namespace.c::mnt_init().

This ensures that sysfs is initialized very early, before anything can
possibily registered with it.

It happens early enough that all filesystems can registered with sysfs
(besides sysfs itself), which is coming next.

23 years agodriver model: Make initialization explicit.
Patrick Mochel [Tue, 4 Mar 2003 04:13:44 +0000 (22:13 -0600)]
driver model: Make initialization explicit.

- Call driver_init() from init/main.c::do_basic_setup().

This ensures that all the driver model subsystems are initialized before
any drivers or devices can be registered.

It nearly frees up the core and postcore initcall levels, making them
available for other kernel code to use freely.

23 years ago[PATCH] IEEE1394 ioctl allocation documentation
Ben Collins [Tue, 4 Mar 2003 02:58:48 +0000 (18:58 -0800)]
[PATCH] IEEE1394 ioctl allocation documentation

23 years agosysfs: fix warning.
Patrick Mochel [Tue, 4 Mar 2003 02:57:38 +0000 (20:57 -0600)]
sysfs: fix warning.

From Andrew Morton

23 years ago[PATCH] IEEE1394 updates
Ben Collins [Tue, 4 Mar 2003 02:55:45 +0000 (18:55 -0800)]
[PATCH] IEEE1394 updates

- Add ioctl32 compat function registration for video1394, dv1394 and
  amdtp.
- Convert SBP-2 driver to new SCSI hotplug host/device interfaces.
- Initial Async ISO (GASP) support.

23 years agopnp interface.c needs <linux/slab.h> for 'kfree()'
Linus Torvalds [Tue, 4 Mar 2003 01:13:00 +0000 (17:13 -0800)]
pnp interface.c needs <linux/slab.h> for 'kfree()'

23 years agoMerge bk://kernel.bkbits.net/davem/sparc-2.5
Linus Torvalds [Tue, 4 Mar 2003 00:42:43 +0000 (16:42 -0800)]
Merge bk://kernel.bkbits.net/davem/sparc-2.5
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years ago[PATCH] sysfs build fix
Andrew Morton [Tue, 4 Mar 2003 00:26:42 +0000 (16:26 -0800)]
[PATCH] sysfs build fix

23 years ago[HOTPLUG]: drivers/base/hotplug.c needs linux/sched.h and linux/string.h
David S. Miller [Mon, 3 Mar 2003 20:27:42 +0000 (12:27 -0800)]
[HOTPLUG]: drivers/base/hotplug.c needs linux/sched.h and linux/string.h

23 years ago[SYSFS]: fs/sysfs/mount.c needs linux/init.h
David S. Miller [Mon, 3 Mar 2003 20:25:17 +0000 (12:25 -0800)]
[SYSFS]: fs/sysfs/mount.c needs linux/init.h

23 years ago[SPARC64]: Fix SMP boot failures.
Bruce D. Elliott [Mon, 3 Mar 2003 20:06:04 +0000 (12:06 -0800)]
[SPARC64]: Fix SMP boot failures.

23 years agosysfs: fix more merge breakage.
Patrick Mochel [Mon, 3 Mar 2003 11:11:30 +0000 (05:11 -0600)]
sysfs: fix more merge breakage.

23 years agosysfs: fixup merge
Patrick Mochel [Mon, 3 Mar 2003 10:59:28 +0000 (04:59 -0600)]
sysfs: fixup merge

23 years agocpufreq: convert to use new interface code.
Patrick Mochel [Mon, 3 Mar 2003 10:38:53 +0000 (04:38 -0600)]
cpufreq: convert to use new interface code.

- Create and register kobjects when adding a cpu to the cpufreq interface,
  instead of the terminally broken struct intf_data.

ACKed by Dominik Brodowski.

23 years ago[PATCH] Another bitop on boolean in pnpbios
John Levon [Mon, 3 Mar 2003 10:13:49 +0000 (02:13 -0800)]
[PATCH] Another bitop on boolean in pnpbios

23 years agoMerge bk://ldm.bkbits.net/linux-2.5-core
Linus Torvalds [Mon, 3 Mar 2003 10:13:05 +0000 (02:13 -0800)]
Merge bk://ldm.bkbits.net/linux-2.5-core
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years ago[PATCH] small tty irq race fix
Nicolas Pitre [Mon, 3 Mar 2003 10:11:31 +0000 (02:11 -0800)]
[PATCH] small tty irq race fix

23 years agoMerge osdl.org:/home/mochel/src/kernel/devel/linux-2.5-virgin
Patrick Mochel [Mon, 3 Mar 2003 07:32:15 +0000 (01:32 -0600)]
Merge osdl.org:/home/mochel/src/kernel/devel/linux-2.5-virgin
into osdl.org:/home/mochel/src/kernel/devel/linux-2.5-core