Updates to sparc, alpha, ppc64, fbdev, XFS, AGP, kbuild, arm...
Likely the last release by me in a while, but Andrew & co can hold the
fort..
Linus
Summary of changes from v2.5.58 to v2.5.59
============================================
<[email protected]>:
o missing break in amd 486 cpu case
<[email protected]>:
o i8k driver update to i8k-1.13
o i8k driver cleanups
o alsa before oss in Kconfig
Ulrich Drepper <[email protected]>:
o new CPUID bit
Falk Hueffner <[email protected]>:
o [AGP] missing includes on Alpha
Jeff Wiedemeier <[email protected]>:
o Fix marvel irq count computation
Nick Holloway <[email protected]>:
o cpia driver update
<[email protected]>:
o fix cardbus/hotplugging
Andi Kleen <[email protected]>:
o x86_64 update
Andrew Morton <[email protected]>:
o ext3 ino_t removal
o factor free memory into max_sane_readahead()
o fix ext3 memory leak
o hugetlbfs: don't implement read/write file_ops
o Use for_each_task_pid() in do_SAK()
o Create a per-cpu proces counter for /proc reporting
o remove has_stopped_jobs()
Anton Blanchard <[email protected]>:
o pp64: move BUG into asm/bug.h
o ppc64: update comment, we now zero extend all 6 arguments in the
32bit syscall path, from Milton Miller
o ppc64: 2.5 module support, from Rusty
o ppc64: fix build when CONFIG_MODULES=n
o ppc64: move BUG_ILLEGAL_INSTR into asm/bug.h, noted by Milton
Miller
o ppc64: remove old strace hack
o ppc64: remove old signal code, unused on 64bit userspace
o ppc64: Remove code which zero/sign extends arguments 5 and 6, its
done unconditionally now
o ppc64: fix exception handling in socket multiplexer
o ppc64: Temporary workaround for oops during coredump
Bjorn Helgaas <[email protected]>:
o [AGP] factor device command updates
o [AGP] fix old pci_find_capability merge botch
o [AGP] Remove unused var
o [AGP] Print AGP version & mode when programming devices
o [AGP] factor device capability collection
o [AGP] use PCI_AGP_* constants
o [AGP] use pci_find_capability in sworks-agp.c
Christoph Hellwig <[email protected]>:
o more procfs bits for !CONFIG_MMU
o remove more junk from i2c headers
o remove some junk from fs/devfs/Makefile
o remove obsolete kern_umount alias for mntput
o fix intermezzo compilation
o don't include coda_fs_i.h in fs.h
o umode_t changes from Adam's mini-devfs
Christoph Hellwig <[email protected]>:
o [XFS] remove superlous MAXNAMELEN checks
o [XFS] some more rename cleanups
o [XFS] xfs_getattr should be static
o [XFS] remount r/o fixes
o [XFS] update xattr.h copyright date
o [XFS] add dmapi miscdevice minor number
o [XFS] fix namespace pullution
o stale bdev reference in quotactl
o fix signed/unsigned issue in SGI partitioning code
o remove GET_USE_COUNT
o remove MOD_IN_USE
Dave Jones <[email protected]>:
o [WATCHDOG] clean up includes
o [WATCHDOG] Final 2.4 bits for advantechwdt
o [WATCHDOG] Final 2.4 bits for eurotechwdt
o [WATCHDOG] Final 2.4 bits for ib700wdt
o [WATCHDOG] Final 2.4 bits for softdog.c
o [WATCHDOG] Final 2.4 changes for w83877f_wdt.c
o [WATCHDOG] final 2.4 fixes for wdt.c
o [WATCHDOG] Final 2.4 changes for wdt285.c
o [WATCHDOG] Final 2.4 changes for wdt_pci.c
o [AGPGART] warning fixes from Bjorn's last patches
o [AGPGART] implement module locking that works
o [AGPGART] Remove ancient unused bits from headers
David Brownell <[email protected]>:
o maintain hcd_dev queue in FIFO order
David S. Miller <[email protected]>:
o [SUNSAB]: Fix uart_get_baud_rate args
o [SPARC64]: Define PAGE_BUG in asm/bug.h
o [SPARC64]: Add UltraSPARC-III cpu frequency driver
o [SPARC64]: Move topology_init to setup.c, it is not SMP specific
o [SPARC64]: Use init/exit facility of cpufreq infrastructure
o [SPARC64]: Update defconfig
Dominik Brodowski <[email protected]>:
o cpufreq: fix compilation, name of gx-suspmod driver
Duncan Sands <[email protected]>:
o USB: kill speedtouch tasklet when shutdown
o USB: make more speedtouch functions static
o USB: SpeedTouch not Speed Touch
Eric Sandeen <[email protected]>:
o [XFS] Make sure we don't walk off the end of the err_level array
o [XFS] Fix dyslexic definition of XFS_MAX_ERR_LEVEL
o [XFS] Merge max file offset fix - use standard Linux macros
o [XFS] Handle mode 0 inodes that find their way onto the unlinked
list These shouldn't be there, probably the result of corruption.
Gabriel Paubert <[email protected]>:
o Cleanup of the lcall7/lcall27 entry path
Geert Uytterhoeven <[email protected]>:
o Amiga keyboard fix
o Q40/Q60 IRQ updates from 2.4.x
o M68k exception table updates
o Sun-3: Add missing deactivate_mm()
o M68k generic RTC driver updates
o Atari ST-RAM swap update
o Q40/Q60 keyboard fixes
o Generic RTC driver documentation
o Mac/m68k NCR5380 SCSI updates
Greg Kroah-Hartman <[email protected]>:
o USB: put the usb storage's SCSI device in the proper place in sysfs
o TTY: add module reference counting for tty drivers
o USB: add dev attribute for usb-serial devices in sysfs
o USB: added .owner for USB drivers that have a struct tty_driver
Greg Ungerer <[email protected]>:
o bug.h for m68knommu arch
o remove BUG from m68knommu arch page.h
o remove obsolete himem.ld from m68knommu sub-arch
o clean up linker symbols in 68EZ328 ucsimm target
o clean up linker symbols in 68EZ328 ucdimm target
o move common macros into m68knommu entry.h
o remove common code from m68knommu/5307 entry.S
o remove common code from m68knommu/68328 entry.S
o remove common code from m68knommu/68360 entry.S
o build common m68knommu entry.S
Henning Meier-Geinitz <[email protected]>:
o Change maintainership of USB scanner driver
Ivan Kokshaysky <[email protected]>:
o alpha ksyms From [email protected]:
o alpha bootp target From [email protected]:
o alpha ipi timeout From [email protected]:
o alpha HARDIRQ_BITS From [email protected]:
o alpha kernel layout From [email protected]:
o alpha osf_shmat lock From [email protected]:
o alpha ev6/ev7 virt_to_phys From [email protected]:
o alpha console callbacks From [email protected]:
o alpha ide hwifs From [email protected]:
o alpha mem_size_limit From [email protected]:
o alpha numa iommu From [email protected]:
o alpha numa update From [email protected]:
o alpha smp fixes From [email protected]:
o alpha kernel start address From [email protected]:
o alpha PCI setup update
o alpha_remap_area_pages From [email protected]:
o alpha titan update From [email protected]:
o alpha irq proc update From [email protected]:
o alpha smp callin From [email protected]:
James Morris <[email protected]>:
o [CRYPTO]: Add support for SHA-386 and SHA-512
o [CRYPTO] remove superfluous goto from des module init exception
path
o [CRYPTO] Add AES and MD4 to tcrypto crypto_alg_available() test
James Simmons <jsimmons@kozmo.(none)>:
o [ATY] Somehow a merge mistake happened. We removed fb_set_var
James Simmons <[email protected]>:
o I810 fbdev updates. Cursor fix for ati mach 64 cards on big endian
machines. Buffer over flow fix for fbcon putcs function. C99
initializers for the STI console drivers.Voodoo 1/2 and NVIDIA
driver updates
o Added resize support for the framebuffer console. Now you can
change the console size via stty. Also support for color palette
changing on VC switch is supported
o [RIVA FBDEV] Driver now uses its own fb_open and fb_release
function again. It has no ill effects. The drivers uses strickly
hardware acceleration so we don't need cfb_fillrect and
cfb_copyarea
o Updates from Helge Deller for the console/fbdev drivers for the
PARISC platform. Small fix for clearing the screen and a string
typo for the Voodoo 1/2 driver
o [MONITOR support] GTF support for VESA complaint monitors. Here we
calculate the general timings needed so we don't over step the
bounds for a monitor
o Remove fb_set_var. Some how it was missed in a merge conflict
o Final updtes to the GTF code. Now the code can gnerate GTF timings
regardless of the validity of info->monospecs
o [TRIDENT FBDEV] Driver ported to the new api
o [GENERIC IMAGEBLIT ACCEL]
o [STI] Updates to latest PARISC changes. Use the latest PCI ids
Kai Germaschewski <[email protected]>:
o Module Sanity Check
o ISDN/HiSax: Fix typo in drivers/isdn/hisax/config.c
o ISDN/HiSax: Fix PnP merge
o ISDN: Fix the janitor fix
o Consolidate read-only sections in arch/*/vmlinux.lds.S
o kbuild: fix broken kallsyms on non-x86 archs
o kbuild/modules: Save space on symbol list
o kbuild: Make asm-generic/vmlinux.lds.h usable for IA-64
o kbuild: kallsyms cleanup
Laszlo Valko <[email protected]>:
o [SPARC64]: Handle SO_TIMESTAMP properly in compat recvmsg
Linus Torvalds <[email protected]>:
o We need to assign resources to cardbus cards _regardless_ of
whether probing tells us they already have a range. The old
information is stale.
o Fix backslash at end of file
o Fix page_address() to not re-evaluate its arguments multiple times
under certain circumstances.
Marc Zyngier <[email protected]>:
o sysfs EISA support
o EISA naming database
o EISA sysfs updates to 3c509 and 3c59x drivers
o EISA sysfs AIP update
Martin J. Bligh <[email protected]>:
o Fix interrupt dest mode / delivery mode confusion
o Add ACPI hook, rename raw_phys_apicid to bios_cpu_apicid
o Make IRQ balancing work with clustered APICs
o Fix APIC header defines for Summit
o Enable Summit in makefile, update summit subarch code
o make vm_enough_memory more efficient
o (1/3) Minimal NUMA scheduler
o (2/3) Initial load balancing
o (3/3) NUMA rebalancer
Matthew Wilcox <[email protected]>:
o acpi_bus_register_driver patch
Nathan Scott <[email protected]>:
o [XFS] Fix up some comments, tidy up some macros - no functional
changes
Patrick Mansfield <[email protected]>:
o USB storage sysfs fix
Patrick Mochel <[email protected]>:
o driver model: update documentation
o kobject: export kset_find_obj
o sysfs: fixup some remaining s390 files
o sysfs: fixup NUMA file that was missed
o sysfs: minor documentation update
o sysfs: fixup SCSI debug driver files
o deadline iosched: make sure queue is valid before unregistering it
o driver model: fix bogus driver binding error reporting and handling
Paul Mackerras <[email protected]>:
o PPC32: Add support for PPC 4xx on-chip devices using the generic
device model.
o PPC32: Page-align the data section of the boot wrapper
o PPC32: Better support for PPC 4xx debug facilities
o PPC32: Use a per-cpu variable for prof_counter and prof_multiplier
Randy Dunlap <[email protected]>:
o update LOG BUF SIZE config
Richard Henderson <[email protected]>:
o [ALPHA] Expose shifts in virt_to_phys to the compiler
o [ALPHA] Use direct calls to titan_ioremap/unmap when building a
titan specific kernel.
o [ALPHA] AGP infrastructure for AGP implemented in Alpha corelogic
(Titan / Marvel), Kconfig and headers.
o [ALPHA] Marvel (AlphaServer ES47, ES80, GS1280) support
o [ALPHA] Fixups to Marvel and Titan for incomplete merging
o [ALPHA] Formatting cleanup, warning removal, move declarations to
header files where they belong.
o [ALPHA] Correct io.h exports and inlining for marvel and titan
o [ALPHA] Corrections to recent vmlinux.lds.S changes
Roger Luethi <[email protected]>:
o export skb_pad symbol
o Fix via-rhine using skb_padto
Roland Dreier <[email protected]>:
o [NET]: Fix up RTM_SETLINK handling
Russell Cattelan <[email protected]>:
o [XFS] Fix the cmn_err stuff to mask the error level before it
checks for max value
o [XFS] make *cmn_err interrupt safe
o [XFS] Revisit the remount read only code again
Russell King <[email protected]>:
o [ARM] Add new system call entries
o [ARM] Remove redundant definitions from ide.h
o [ARM] Fix CPUFREQ initialisation oops
o [ARM] Update sa1100fb
o [ARM] Update acornfb for new fbcon layer
o [ARM] Use new asm/bug.h for arch/arm/kernel/bios32.c
o [ARM] Prevent "scheduling while atomic" in cpu_idle()
o [ARM] Update mach-types; add 8 new machine types, fix karo entry
o [ARM] Fix failure paths in fd1772.c initialisation
o [ARM/IDE] Fix BLK_DEV_IDEDMA setting on non-Acorn ARM systems
o [ARM] Fix Integrator __virt_to_bus/__bus_to_virt
Stephen Rothwell <[email protected]>:
o compat_{old_}sigset_t generic part
o compat_{old_}sigset_t s390x part
o compat_sys_sigpending and compat_sys_sigprocmask
o compat_sys_sigpending and compat_sys_sigprocmask
Tom Rini <[email protected]>:
o PPC32: Change the MontaVista copyright / GPL boilerplate to a
condensed version.
Trond Myklebust <[email protected]>:
o Fix RPC client warning in 2.5.58
o Fix NFS root mount handling
Hey all,
Just done a bk pull and got this
# make mrproper; make menuconfig
....
gcc -o scripts/lxdialog/lxdialog scripts/lxdialog/checklist.o
scripts/lxdialog/menubox.o scripts/lxdialog/textbox.o
scripts/lxdialog/yesno.o scripts/lxdialog/inputbox.o scripts/lxdialog/util.o
scripts/lxdialog/lxdialog.o scripts/lxdialog/msgbox.o -lncurses
./scripts/kconfig/mconf arch/i386/Kconfig
arch/i386/Kconfig:1185: can't open file "drivers/eisa/Kconfig"
make: *** [menuconfig] Error 1
Carl
On Thu, Jan 16, 2003 at 10:12:23PM -0600, Kai Germaschewski wrote:
> > ./scripts/kconfig/mconf arch/i386/Kconfig
> > arch/i386/Kconfig:1185: can't open file "drivers/eisa/Kconfig"
> > make: *** [menuconfig] Error 1
>
> bk -r get -q
>
> or just
>
> bk get drivers/eisa
>
> in this case. I guess this is becoming a FAQ.
It would be cool if the the Makefile let make knew about these dependencies
so they would be checked out automagically.
--
Anders Gustafsson - [email protected] - http://0x63.nu/
On Fri, 17 Jan 2003, Carl Gherardi wrote:
| Hey all,
|
| Just done a bk pull and got this
|
| # make mrproper; make menuconfig
| ....
| gcc -o scripts/lxdialog/lxdialog scripts/lxdialog/checklist.o
| scripts/lxdialog/menubox.o scripts/lxdialog/textbox.o
| scripts/lxdialog/yesno.o scripts/lxdialog/inputbox.o scripts/lxdialog/util.o
| scripts/lxdialog/lxdialog.o scripts/lxdialog/msgbox.o -lncurses
| ./scripts/kconfig/mconf arch/i386/Kconfig
| arch/i386/Kconfig:1185: can't open file "drivers/eisa/Kconfig"
| make: *** [menuconfig] Error 1
Hm, that's odd. That file is in the downloaded .bz2 file.
I would pull again or do 'bk changes -R' or some other bk
command to check its integrity.
Or maybe it's just not 'got' (checked out) by bk...
--
~Randy
On Fri, 17 Jan 2003, Carl Gherardi wrote:
> Just done a bk pull and got this
>
> # make mrproper; make menuconfig
> ....
> gcc -o scripts/lxdialog/lxdialog scripts/lxdialog/checklist.o
> scripts/lxdialog/menubox.o scripts/lxdialog/textbox.o
> scripts/lxdialog/yesno.o scripts/lxdialog/inputbox.o scripts/lxdialog/util.o
> scripts/lxdialog/lxdialog.o scripts/lxdialog/msgbox.o -lncurses
> ./scripts/kconfig/mconf arch/i386/Kconfig
> arch/i386/Kconfig:1185: can't open file "drivers/eisa/Kconfig"
> make: *** [menuconfig] Error 1
bk -r get -q
or just
bk get drivers/eisa
in this case. I guess this is becoming a FAQ.
--Kai
On Fri, Jan 17, 2003 at 05:16:28AM +0100, Anders Gustafsson wrote:
> It would be cool if the the Makefile let make knew about these dependencies
> so they would be checked out automagically.
Indeed. Couldn't agree more.
--
---
Larry McVoy lm at bitmover.com http://www.bitmover.com/lm
Carl Gherardi wrote:
> Hey all,
>
> Just done a bk pull and got this
>
> # make mrproper; make menuconfig
> ....
> gcc -o scripts/lxdialog/lxdialog scripts/lxdialog/checklist.o
> scripts/lxdialog/menubox.o scripts/lxdialog/textbox.o
> scripts/lxdialog/yesno.o scripts/lxdialog/inputbox.o scripts/lxdialog/util.o
> scripts/lxdialog/lxdialog.o scripts/lxdialog/msgbox.o -lncurses
> ./scripts/kconfig/mconf arch/i386/Kconfig
> arch/i386/Kconfig:1185: can't open file "drivers/eisa/Kconfig"
> make: *** [menuconfig] Error 1
You need to check that file out.
After every run of "bk pull", I always do
bk -r co -Sq
On Fri, 17 Jan 2003, Anders Gustafsson wrote:
> On Thu, Jan 16, 2003 at 10:12:23PM -0600, Kai Germaschewski wrote:
> > > ./scripts/kconfig/mconf arch/i386/Kconfig
> > > arch/i386/Kconfig:1185: can't open file "drivers/eisa/Kconfig"
> > > make: *** [menuconfig] Error 1
> >
> > bk -r get -q
> >
> > or just
> >
> > bk get drivers/eisa
> >
> > in this case. I guess this is becoming a FAQ.
>
> It would be cool if the the Makefile let make knew about these dependencies
> so they would be checked out automagically.
Unfortunately, the Makefile doesn't really know about the Kconfig files,
the "source drivers/whatever/Kconfig" commands are in Kconfig, and
duplicating them into the Makefile would be rather error-prone.
Even if that was done, the Makefiles also cannot know about e.g. headers
included into C files, so it'd die at that point. At some point I hacked a
LD_PRELOAD library which would try to exec a "get" when open(2) fails,
which fixes gcc, kconfig and whatnotsoever. I suppose a better solution is
"checkout: get", though.
--Kai
On Thu, Jan 16, 2003 at 08:17:39PM -0800, Larry McVoy wrote:
> A little know option which makes things go faster is
>
> bk -r get -qS
>
> which gets only those files not already gotten. Linus has asked why this
> isn't the default and the only reason I can give him is that it is an
> interface change and we'll do it in bk 4.0. It's the right answer.
Whats the difference between bk co and bk get?
--
Anders Gustafsson - [email protected] - http://0x63.nu/
A little know option which makes things go faster is
bk -r get -qS
which gets only those files not already gotten. Linus has asked why this
isn't the default and the only reason I can give him is that it is an
interface change and we'll do it in bk 4.0. It's the right answer.
> bk -r get -q
>
> or just
>
> bk get drivers/eisa
>
> in this case. I guess this is becoming a FAQ.
>
> --Kai
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
---
Larry McVoy lm at bitmover.com http://www.bitmover.com/lm
> Whats the difference between bk co and bk get?
co has RCS compat options and get has SCCS compat options. If you are
used to CVS/RCS, "bk co" and "bk ci" are your friend, if you are used
to SCCS then "bk get" and "bk delta" are what you like. They do the
same thing.
--
---
Larry McVoy lm at bitmover.com http://www.bitmover.com/lm
On Thu, Jan 16, 2003 at 06:28:03PM -0800, Linus Torvalds wrote:
> Updates to sparc, alpha, ppc64, fbdev, XFS, AGP, kbuild, arm...
> Likely the last release by me in a while, but Andrew & co can hold the
> fort..
struct thread_info is shared with the stack, not struct task_struct.
False positives have been seen.
-- wli
===== arch/i386/kernel/irq.c 1.24 vs edited =====
--- 1.24/arch/i386/kernel/irq.c Thu Oct 31 07:28:34 2002
+++ edited/arch/i386/kernel/irq.c Thu Jan 16 20:39:53 2003
@@ -338,9 +338,9 @@
__asm__ __volatile__("andl %%esp,%0" :
"=r" (esp) : "0" (8191));
- if (unlikely(esp < (sizeof(struct task_struct) + 1024))) {
+ if (unlikely(esp < (sizeof(struct thread_info) + 1024))) {
printk("do_IRQ: stack overflow: %ld\n",
- esp - sizeof(struct task_struct));
+ esp - sizeof(struct thread_info));
dump_stack();
}
}
Doh! This is something I've always wanted to know. Do you know how long
bk -r get takes on a slow system with a badly implemented IDE driver? :(
Well, now I know. Great :D
Regards
Josh
On Thu, Jan 16, 2003 at 08:17:39PM -0800, Larry McVoy wrote:
> A little know option which makes things go faster is
>
> bk -r get -qS
>
> which gets only those files not already gotten. Linus has asked why this
> isn't the default and the only reason I can give him is that it is an
> interface change and we'll do it in bk 4.0. It's the right answer.
>
> > bk -r get -q
> >
> > or just
> >
> > bk get drivers/eisa
> >
> > in this case. I guess this is becoming a FAQ.
> >
> > --Kai
> >
> >
> > -
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to [email protected]
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at http://www.tux.org/lkml/
>
> --
> ---
> Larry McVoy lm at bitmover.com http://www.bitmover.com/lm
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
.-`-.-`-.-=============----->
Joshua Kwan [email protected]
[email protected]
On Thu, 16 Jan 2003 18:28:03 -0800 (PST) Linus Torvalds (LT) wrote:
LT> Updates to sparc, alpha, ppc64, fbdev, XFS, AGP, kbuild, arm...
LT> James Simmons <[email protected]>:
LT> o [RIVA FBDEV] Driver now uses its own fb_open and fb_release
LT> function again. It has no ill effects. The drivers uses strickly
LT> hardware acceleration so we don't need cfb_fillrect and
LT> cfb_copyarea
Hi James,
The 2.5.59 version of the rivafb driver has some strange effects on my screen.
The leftmost ~5 pixels of each line are displaced to the right side of the screen,
so that the remainder of the screen is shifted ~5 pixels left, i.e. it's like
the screen has been rotated left a bit. 2.5.58 was alright.
If you've got a patch for this issue, I'll test it.
Regards,
-Udo.
Hi James,
Here's some other info from boot messages that suggests something is going wrong.
rivafb: nVidia device/chipset 10DE0150
rivafb: RIVA MTRR set to ON
rivafb: PCI nVidia NV10 framebuffer ver 0.9.5b (nVidiaGeForce2-G, 32MB @ 0xC8000000)
Badness in kobject_register at lib/kobject.c:152
Call Trace:
[<c0209ef8>] kobject_register+0x58/0x70
[<c024532b>] bus_add_driver+0x5b/0xe0
[<c024579f>] driver_register+0x2f/0x40
[<c0176208>] create_proc_entry+0x88/0xd0
[<c020fe37>] pci_register_driver+0x47/0x60
[<c010507a>] init+0x3a/0x160
[<c0105040>] init+0x0/0x160
[<c010713d>] kernel_thread_helper+0x5/0x18
Console: switching to colour frame buffer device 144x54
Regards,
-Udo.
[email protected] said:
> A little know option which makes things go faster is
> bk -r get -qS
> which gets only those files not already gotten. Linus has asked why
> this isn't the default and the only reason I can give him is that it
> is an interface change and we'll do it in bk 4.0. It's the right
> answer.
Isn't there some way to tell BK to extract the files _while_ it's pulling
the deltas. You know; while it's all right there in the cache?
--
dwmw2
On Thu, Jan 16, 2003 at 06:28:03PM -0800, Linus Torvalds wrote:
>...
> Summary of changes from v2.5.58 to v2.5.59
> ============================================
>...
> Marc Zyngier <[email protected]>:
>...
> o EISA sysfs updates to 3c509 and 3c59x drivers
>...
This change browke the compilation of 3c509 with CONFIG_PM:
<-- snip -->
...
gcc -Wp,-MD,drivers/net/.3c509.o.d -D__KERNEL__ -Iinclude -Wall
-Wstrict-prototypes -Wno-trigraphs -O2 -fno-strict-aliasing -fno-common
-pipe -mpreferred-stack-boundary=2 -march=k6
-Iinclude/asm-i386/mach-default -nostdinc -iwithprefix include
-DKBUILD_BASENAME=3c509 -DKBUILD_MODNAME=3c509 -c -o
drivers/net/3c509.o drivers/net/3c509.c
drivers/net/3c509.c: In function `el3_common_init':
drivers/net/3c509.c:324: `card_idx' undeclared (first use in this function)
drivers/net/3c509.c:324: (Each undeclared identifier is reported only once
drivers/net/3c509.c:324: for each function it appears in.)
drivers/net/3c509.c: In function `el3_probe':
drivers/net/3c509.c:360: warning: `dev' might be used uninitialized in this function
drivers/net/3c509.c: At top level:
drivers/net/3c509.c:268: warning: `nopnp' defined but not used
make[2]: *** [drivers/net/3c509.o] Error 1
<-- snip -->
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
>>>>> "Adrian" == Adrian Bunk <[email protected]> writes:
Adrian> On Thu, Jan 16, 2003 at 06:28:03PM -0800, Linus Torvalds wrote:
>> ...
>> o EISA sysfs updates to 3c509 and 3c59x drivers
>> ...
Adrian> This change browke the compilation of 3c509 with CONFIG_PM:
Can you try this patch (compiles, but otherwise untested) ?
Thanks,
M.
===== drivers/net/3c509.c 1.30 vs edited =====
--- 1.30/drivers/net/3c509.c Wed Jan 15 11:07:35 2003
+++ edited/drivers/net/3c509.c Fri Jan 17 11:17:18 2003
@@ -319,16 +319,6 @@
dev->watchdog_timeo = TX_TIMEOUT;
dev->do_ioctl = netdev_ioctl;
-#ifdef CONFIG_PM
- /* register power management */
- lp->pmdev = pm_register(PM_ISA_DEV, card_idx, el3_pm_callback);
- if (lp->pmdev) {
- struct pm_dev *p;
- p = lp->pmdev;
- p->data = (struct net_device *)dev;
- }
-#endif
-
return 0;
}
@@ -597,6 +587,16 @@
lp->pnpdev = idev;
#endif
lp->mca_slot = mca_slot;
+
+#ifdef CONFIG_PM
+ /* register power management */
+ lp->pmdev = pm_register(PM_ISA_DEV, card_idx, el3_pm_callback);
+ if (lp->pmdev) {
+ struct pm_dev *p;
+ p = lp->pmdev;
+ p->data = (struct net_device *)dev;
+ }
+#endif
return el3_common_init (dev);
}
--
Places change, faces change. Life is so very strange.
On Fri, Jan 17, 2003 at 11:23:28AM +0100, Marc Zyngier wrote:
> >>>>> "Adrian" == Adrian Bunk <[email protected]> writes:
>
> Adrian> On Thu, Jan 16, 2003 at 06:28:03PM -0800, Linus Torvalds wrote:
> >> ...
> >> o EISA sysfs updates to 3c509 and 3c59x drivers
> >> ...
>
> Adrian> This change browke the compilation of 3c509 with CONFIG_PM:
>
> Can you try this patch (compiles, but otherwise untested) ?
Since I'm doing compile-only tests but don't have the hardware available
all I can do is to confirm that this patch fixes the compilation.
> Thanks,
>
> M.
>...
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
On Thu, 16 Jan 2003, Kai Germaschewski wrote:
> On Fri, 17 Jan 2003, Anders Gustafsson wrote:
> > On Thu, Jan 16, 2003 at 10:12:23PM -0600, Kai Germaschewski wrote:
> > > > ./scripts/kconfig/mconf arch/i386/Kconfig
> > > > arch/i386/Kconfig:1185: can't open file "drivers/eisa/Kconfig"
> > > > make: *** [menuconfig] Error 1
> > >
> > > bk -r get -q
> > >
> > > or just
> > >
> > > bk get drivers/eisa
> > >
> > > in this case. I guess this is becoming a FAQ.
> >
> > It would be cool if the the Makefile let make knew about these dependencies
> > so they would be checked out automagically.
>
> Unfortunately, the Makefile doesn't really know about the Kconfig files,
> the "source drivers/whatever/Kconfig" commands are in Kconfig, and
> duplicating them into the Makefile would be rather error-prone.
What about learning `make depend' a bit Kconfig syntax?
> Even if that was done, the Makefiles also cannot know about e.g. headers
> included into C files, so it'd die at that point. At some point I hacked a
> LD_PRELOAD library which would try to exec a "get" when open(2) fails,
> which fixes gcc, kconfig and whatnotsoever. I suppose a better solution is
> "checkout: get", though.
Isn't all of this in .depend?
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On Fri, 17 Jan 2003, Geert Uytterhoeven wrote:
> > Unfortunately, the Makefile doesn't really know about the Kconfig files,
> > the "source drivers/whatever/Kconfig" commands are in Kconfig, and
> > duplicating them into the Makefile would be rather error-prone.
>
> What about learning `make depend' a bit Kconfig syntax?
"make depend" is basically gone.
> > Even if that was done, the Makefiles also cannot know about e.g. headers
> > included into C files, so it'd die at that point. At some point I hacked a
> > LD_PRELOAD library which would try to exec a "get" when open(2) fails,
> > which fixes gcc, kconfig and whatnotsoever. I suppose a better solution is
> > "checkout: get", though.
>
> Isn't all of this in .depend?
In 2.5 things work differently, so no.
--Kai
Compile statistics: 2.5.59
Not much change.
2.5.58 2.5.59
------------------------ ------------------------
bzImage (defconfig) 20 warnings/0 errors 20 warnings/0 errors
bzImage (allmodconfig) 32 warnings/9 errors 32 warnings/9 errors
modules (allmodconfig) 3156 warnings/154 errors 3119 warnings/159 errors
Compile statistics have been for kernel releases from 2.5.46 to 2.5.59
at: http://www.osdl.org/archive/cherry/stability
John
> Compile statistics: 2.5.59
>
> Not much change.
>
> 2.5.58 2.5.59
> ------------------------ ------------------------
> bzImage (defconfig) 20 warnings/0 errors 20 warnings/0 errors
> bzImage (allmodconfig) 32 warnings/9 errors 32 warnings/9 errors
> modules (allmodconfig) 3156 warnings/154 errors 3119 warnings/159 errors
>
> Compile statistics have been for kernel releases from 2.5.46 to 2.5.59
> at: http://www.osdl.org/archive/cherry/stability
I think tracking this this a great idea - all these warnings make it really
hard to see what's going on. Is there any change you could add a PAE enabled
config to the setup? Tends to generate lots of stupid warnings about
typecasts when people do printk("%08lx", dma_addr_t_thingy);
Thanks,
M.
On Fri, Jan 17, 2003 at 08:55:16AM -0800, John Cherry wrote:
> Compile statistics: 2.5.59
>
> Not much change.
>
> 2.5.58 2.5.59
> ------------------------ ------------------------
> bzImage (defconfig) 20 warnings/0 errors 20 warnings/0 errors
> bzImage (allmodconfig) 32 warnings/9 errors 32 warnings/9 errors
> modules (allmodconfig) 3156 warnings/154 errors 3119 warnings/159 errors
>
> Compile statistics have been for kernel releases from 2.5.46 to 2.5.59
> at: http://www.osdl.org/archive/cherry/stability
Have a couple of extra warnings:
fs/binfmt_elf.c: In function `create_elf_tables':
fs/binfmt_elf.c:239: warning: initialization makes integer from pointer without a cast
fs/binfmt_elf.c:249: warning: initialization makes integer from pointer without a cast
#ifndef elf_addr_t
#define elf_addr_t unsigned long
#endif
elf_addr_t *argv, *envp;
__put_user(NULL, argv);
__put_user(NULL, envp);
It would therefore appear that x86 __put_user is not properly type-checking
the arguments to __put_user().
Here's a patch which fixes the warning (but doesn't fix x86's type-check
challenged __put_user implementation):
--- orig/fs/binfmt_elf.c Thu Nov 28 16:45:26 2002
+++ linux/fs/binfmt_elf.c Fri Jan 17 17:08:50 2003
@@ -236,7 +236,7 @@
return;
p += len;
}
- __put_user(NULL, argv);
+ __put_user(0, argv);
current->mm->arg_end = current->mm->env_start = p;
while (envc-- > 0) {
size_t len;
@@ -246,7 +246,7 @@
return;
p += len;
}
- __put_user(NULL, envp);
+ __put_user(0, envp);
current->mm->env_end = p;
/* Put the elf_info on the stack in the right place. */
--
Russell King ([email protected]) The developer of ARM Linux
http://www.arm.linux.org.uk/personal/aboutme.html
On Fri, Jan 17, 2003 at 10:36:54AM -0600, Kai Germaschewski wrote:
> > Isn't all of this in .depend?
>
> In 2.5 things work differently, so no.
The longer answer, if you feel bored :-)
In 2.5 kbuild generate a file that list all dependencies:
.foo.o.cmd
This file is generated the first time the file is compiled, and contains
dependencies too all included files + all referred CONFIG_* options.
What to keep in mind is that make does NOT know the dependencies when run
the very first time. Thats because there is no need to generate the
.foo.o.cmd file when the .o file does not exist - make will build it no
matter what.
Therefore if foo.c depends on <linux/whatever.h> make will not
see this dependency, and gcc will complain.
Though - at the second run make will know the dependency and 'get' the file.
Kconfig files are special in the sense that we do not always check
the full dependency on them.
We have autoconf.h, config/MARKER and more magic involved here.
Since kconfig does not generate a .config.cmd when it fails to locate
a Kconfig file, it would not help to clean up this - with respect to
BK/CVS integration.
HTH,
Sam - kbuild apprentice ;-)
On Freitag, 17. Januar 2003 17:55, John Cherry wrote:
...
> Compile statistics have been for kernel releases from 2.5.46 to 2.5.59
> at: http://www.osdl.org/archive/cherry/stability
think on old oldconfigs:
--- compregress.sh.old 2003-01-18 04:49:26.000000000 +0100
+++ compregress.sh 2003-01-18 04:51:17.000000000 +0100
@@ -231,18 +231,16 @@
if [ $HAS_OLDCONFIG == 1 ]; then
printf " Making bzImage (oldconfig): "
+ STR="\n"
for x in 1 2 3 4 5 6 7 8 9 10; do
- echo "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n" >> tmp_return
+ STR="$STR$STR"
done
- STR=`cat tmp_return`
- rm -f tmp_return
test -f .config && cp -f .config .config.bak
make mrproper > /dev/null 2>&1
test -f .config.bak && mv .config.bak .config
echo -e $STR | make oldconfig &> /dev/null
- echo -e $STR | make oldconfig &> /dev/null
make dep >> $KERNEL_OLDCONFIG 2>&1
make $MAKEOPT bzImage >> $KERNEL_OLDCONFIG 2>&1
WARN_COUNT=`egrep "warning:" $KERNEL_OLDCONFIG | wc -l`
> Tom Rini <[email protected]>:
> o PPC32: Change the MontaVista copyright / GPL boilerplate to a
> condensed version.
Where did you find the GPLv2.1 these new boilerplates refer to?