ChangeSet@1.1083, 2003-08-26 16:05:07-03:00, johnstul@us.ibm.com [PATCH] Handle clustered XAPIC in set_ioapic_affinity() Marcelo, All, In set_ioapic_affinity() we take a cpu mask which we normally pass into the apic. However, on xapics in clustered physical mode (ie: x440s) we need to convert this mask into a physical apicid before sending it to the apic. If clustered_apic_mode is XAPIC, this patch selects the lowest cpu in the mask and routes the interrupts there. This avoids timedoubling and other interrupt misdirection seen on x440s after playing with /proc/irq/x/smp_affinity. Please apply. thanks -john ChangeSet@1.1082, 2003-08-26 16:01:48-03:00, stelian@popies.net [PATCH] meye driver updates The attached patch updates the meye driver with some tiny changes: * use SetPageReserved instead of mem_map_reserve, and remove the wrapper.h dependency * document the supported hardware. ===== Documentation/video4linux/meye.txt 1.6 vs edited ===== ChangeSet@1.1081, 2003-08-26 16:01:08-03:00, stelian@popies.net [PATCH] sonypi driver update Hi, Now that we are starting a new -pre cycle, here are the latest sonypi updates: * document the fact that FX501/FX702 laptops are not supported * add battery insert/remove events (thanks to Daniel K.) * improve the event detection using a different port offset on 'type2' models (thanks to Daniel K.) * now that ACPI is finally in the kernel, simplify the ACPI tests. Marcelo, please apply. Stelian. ===== Documentation/sonypi.txt 1.12 vs edited ===== ChangeSet@1.1080, 2003-08-26 15:37:53-03:00, hch@lst.de [PATCH] reserve a sysctl number for XFS (pagebuf) Reserve the 2.5 number for VM_PAGEBUF in 2.4 aswell. ChangeSet@1.1079, 2003-08-26 15:29:58-03:00, johnstul@us.ibm.com [PATCH] Do not clear SMI pin at bootup Marcelo, All, Some of our more recent hardware requires that SMIs are routed through the IOAPIC, thus when we clear_IO_APIC() at boot time, we clear the BIOS initialized SMI pin (which is unfortunately not described in the MPS or ACPI tables). This basically clobbers the SMI, which can cause problems with console redirection as well keeping us from being able to transition into full ACPI mode. This patch (back ported from 2.5) simply reads the apic entry in clear_IO_APIC to make sure the delivery_mode isn't dest_SMI. If it is, we leave the apic entry alone and return. With this patch, booting with full ACPI works and SMIs function properly. Please consider for acceptance. thanks -john ChangeSet@1.1078, 2003-08-26 15:28:13-03:00, willy@w.ods.org [PATCH] make log buffer length selectable On Mon, Aug 25, 2003 at 04:48:30AM -0700, Marcelo Tosatti wrote: > final: > > - 2.4.22-rc4 was released as 2.4.22 with no changes. Hi Marcelo, as you requested, here is the log_buf_len patch for inclusion in 23-pre. Cheers, Willy ChangeSet@1.1077, 2003-08-26 11:52:36-03:00, hch@lst.de [PATCH] use list_add_tail in buffer_insert_list This is a really old patch from the XFS tree. We need to call list_add_tail in buffer_insert_list to preserve buffer ordering. This essential for a good extent layout with XFS's delayed allocation and a while ago the reiserfs group requested the same change. ACKed by sct. ChangeSet@1.1076, 2003-08-26 10:38:45-03:00, marcelo@logos.cnet Merge bk://kernel.bkbits.net/davem/net-2.4 into logos.cnet:/home/marcelo/bk/linux-2.4 ChangeSet@1.1075, 2003-08-26 09:51:54-03:00, arjanv@redhat.com [PATCH] Fix asm constraint bug in arch/i386/kernel/pci-pc.c Hi, Reported by HP, found by Jakub. The patch below fixes a missing asm constraint. The asm constraint of this function needs to depend on (opt) so that gcc knows that the initialisation of the opt structure needs to happen before the asm code; in the old code only the address of opt was in the constraints which isn't enough. Newer gcc will just miscompile this for p4. Greetings, Arjan van de Ven ChangeSet@1.1074, 2003-08-26 09:51:29-03:00, ak@muc.de [PATCH] Compile fix for ACPI in 2.4.22/x86-64 Marcelo unfortunately added an last minute ACPI update that changed ACPI interfaces and broke x86-64 compilation. I didn't catch it in time, so 2.4.22 does not compile out of the box for AMD64. This patch fixes it. You'll have to apply it when compiling 2.4.22 for x86-64 -Andi ChangeSet@1.1065.11.1, 2003-08-25 18:56:42-04:00, jgarzik@redhat.com Merge redhat.com:/garz/repo/marcelo-2.4 into redhat.com:/garz/repo/net-drivers-2.4 ChangeSet@1.1065.10.10, 2003-08-25 19:28:53-03:00, mikpe@csd.uu.se [PATCH] 2.4.22 local APIC updates 3/3: disable APIC_BASE on reboot Marcelo, This patch ensures that we properly disable the local APIC before reboot. This fixes BIOS problems reported by a few people. disable_local_APIC() now checks if detect_init_APIC() enabled the local APIC via the APIC_BASE MSR, and if so it now disables APIC_BASE. Previously we would leave APIC_BASE enabled, and that made some BIOSen malfunction. The SMP reboot code calls disable_local_APIC(). On SMP HW there is no change since detect_init_APIC() isn't called and APIC_BASE isn't enabled by us. An SMP kernel on UP HW behaves just like an UP_APIC kernel, so it disables APIC_BASE if we enabled it at boot. The UP_APIC suspend code is simplified since the existing code to disable APIC_BASE is moved into disable_local_APIC(). (Felix Kühling originally reported the BIOS reboot problem. This is a fixed-up version of his preliminary patch.) Backport from 2.6.0-test4. Tested in 2.4.22-rc. Please apply. /Mikael ChangeSet@1.1065.10.9, 2003-08-25 19:28:16-03:00, mikpe@csd.uu.se [PATCH] 2.4.22 local APIC updates 2/3: add lapic/nolapic options Marcelo, This patch adds the lapic and nolapic kernel options, which give users greater control of the local APIC enable process on x86: - "nolapic" prevents the kernel from enabling or using the local APIC. This is stronger than listing a machine in the DMI scan blacklist, since it also works for machines that boot with the local APIC already enabled. "nolapic" can also be seen as a simple-to-deploy alternative to keeping the DMI blacklist rules up-to-date. - "lapic" tells the kernel to force-enable the P4 local APIC if the BIOS disabled it. This is needed on some machines. The default (don't enable if BIOS didn't) can't be changed since there is an even larger number of P4s where ACPI breaks if the local APIC is enabled. Backport from 2.6.0-test4. Tested in 2.4.22-rc. Please apply. This patch renames the dont_enable_local_apic variable, so if you apply it you also need to apply "local APIC updates 1/3: remove incorrect blacklist rules". /Mikael ChangeSet@1.1065.10.8, 2003-08-25 19:26:55-03:00, mikpe@csd.uu.se [PATCH] 2.4.22 local APIC updates 1/3: remove incorrect blacklist rules Marcelo, This patch removes the x86 local APIC blacklist rules for the Microstar 6163 and Intel AL440LX mainboards, as these boards _do_ work with local APIC enabled. The AL440LX blacklist rule was the only place where the dont_enable_local_apic_timer flag could be set, so the patch removes that variable and the code which checks it. The hangs these boards used to have has since been traced to a combination of using specific graphics cards with APM's DISPLAY_BLANK option. Backport from 2.5.72. Also tested in 2.4 for months. Please apply. /Mikael ChangeSet@1.1071, 2003-08-25 18:26:15-04:00, jgarzik@redhat.com Merge redhat.com:/garz/repo/marcelo-2.4 into redhat.com:/garz/repo/hwrandom-2.4 ChangeSet@1.1065.10.7, 2003-08-25 19:21:59-03:00, m.c.p@wolk-project.de [PATCH] Proper APIC with HyperThreading Hi Marcelo, the boot issue with DL760 G2 issue with Hyper-Threading (16-way logical) was root-caused to be coming from hardcoding of apic-version during the acpi boot-up code. The attached patch fixes the issue, by reading in the actual apic version. This apic-version is used later in the boot process, to identify whether the processor supports xAPIC or not. And that in turn was resulting in boot failure. Refer following lkml threads for detailed the problem description: Subject: SUMMARY: DL760 G2 issue with Hyper-Threading: Subject: PROBLEM: DL760 G2 issue on 2.4.21 with Hyper-Threading In -ac and -wolk for some time. Patch by "Venkatesh Pallipadi" ChangeSet@1.1065.10.6, 2003-08-25 19:18:30-03:00, m.c.p@wolk-project.de [PATCH] Avoid potentially leaking pagetables into the per-cpu queues Hi Marcelo, $subject says it all. Patch from Andrea Arcangeli! It's in -aa since 2.4.17rc2aa1, in -wolk for some time. ChangeSet@1.1065.10.5, 2003-08-25 19:11:07-03:00, marcelo@logos.cnet Cset exclude: m.c.p@wolk-project.de|ChangeSet|20030825194257|34486 ChangeSet@1.1065.1.54, 2003-08-25 14:51:42-07:00, davem@nuts.ninka.net [NET]: Include asm/uaccess.h in net/core/ethtool.c ChangeSet@1.1070, 2003-08-25 17:38:58-04:00, jgarzik@redhat.com Merge redhat.com:/garz/repo/marcelo-2.4 into redhat.com:/garz/repo/hwrandom-2.4 ChangeSet@1.1069, 2003-08-25 17:37:39-04:00, jgarzik@redhat.com [ia32] mention that X86_VENDOR_ID is tied to NCAPINTS, in a comment in arch/i386/kernel/head.S. ChangeSet@1.1065.10.4, 2003-08-25 18:24:46-03:00, m.c.p@wolk-project.de [PATCH] CONFIG_NR_CPUS Hi Marcelo, this patch has been around for ages. 2.6 has the same change/cleanup. Now, w/o the patch, the kernel will be compiled with 32 CPUs, allthough you might only have 2 or 4 or so. This is purely to save memory - each supported CPU adds approximately eight kilobytes to the kernel image. ChangeSet@1.1065.1.53, 2003-08-25 14:24:34-07:00, jgarzik@redhat.com [NET]: Backport ethtool_ops from 2.6.x ChangeSet@1.1065.10.3, 2003-08-25 18:18:51-03:00, m.c.p@wolk-project.de [PATCH] LVM Update v1.0.5 to v1.0.7 On Thursday 31 July 2003 03:04, Marcelo Tosatti wrote: Hi Marcelo, > > would you consider updating LVM from 1.05 to 1.07 in -pre10? It's out > > there for many months now, many vendor tree's are using it, my tree uses > > it and prolly many others and I am not aware of any problems. > > Should I send a patch? Please let me know. > It depends on the type of changes. I prefer including it in -pre10 only if > its obvioux bugfixes only. > Please send me it together with a detailed changelog. ok, here we go, 5 small but important bugfixes: Changes from v1.0.5 -> v1.0.7 ----------------------------- o fixed lvrename bug leaving wrong filesystem entries around in /proc/lvm/ o fixed lvextend/lvreduce bug on snapshots preventing correct reload of the excpetion table possibly giving wrong snapshot content. o fixed snapshot extend/reduce memory leak o fixed wrong VG PE counter o some whitespace cleanup ChangeSet@1.1065.10.1, 2003-08-25 18:06:46-03:00, m.c.p@wolk-project.de [PATCH] Handle get_block errors correctly in block_read_full_page() Hi Marcelo, quoting Andrea: > Anton found get_block errors weren't handled correctly by > block_read_full_page(). This is my proposed fix for the problem, > implementation and patching is from Anton Altaparmakov. It's in -aa and -wolk for some time. Please apply for 2.4.23-pre1. Thank you :) ChangeSet@1.1065.1.52, 2003-08-25 14:00:11-07:00, davem@nuts.ninka.net [TG3]: Fix AC1001 typo in pci_ids.h ChangeSet@1.1065.9.1, 2003-08-25 13:41:49-07:00, davem@nuts.ninka.net Merge nuts.ninka.net:/home/davem/src/BK/sparcwork-2.4 into nuts.ninka.net:/home/davem/src/BK/sparc-2.4 ChangeSet@1.1065.1.51, 2003-08-25 13:38:25-07:00, davem@nuts.ninka.net Merge nuts.ninka.net:/home/davem/src/BK/network-2.4 into nuts.ninka.net:/home/davem/src/BK/net-2.4 ChangeSet@1.1065.1.50, 2003-08-25 16:44:39-03:00, m.c.p@wolk-project.de [PATCH] Unblank console if panic() Hi Marcelo, if a kernel panic()'s, and a console blank timer is running, it is possible that the screen goes blank and he/she isn't able to read the panic. Bad. The attached patch disables the console blank timer when panic(). Patch from Daniel Phillips! It's in -wolk for a long time. NOTE: You have to apply "2.4-panic-console-switch-1.patch" first, otherwise this will fail with 1 hunk. ChangeSet@1.1065.1.49, 2003-08-25 16:42:57-03:00, m.c.p@wolk-project.de [PATCH] Allow sysrq() via /proc/sys/kernel/magickey Hi Marcelo, sysrq() is a good thing to debug things, but unfortunately you need to have physical access to the keyboard. My company for instance maintains tons of remote machines and sometimes we need to do sysrq() too, but it's not possible to do so the remote way. Attached patch enables emulation of the Magic SysRq key (Alt-SysRq-key) via the /proc interface. Just echo the desired character into the file and there you go. Patch from Randy Dunlap! It's in -wolk for a long time and also in some other kernel tree forks. 2.5/2.6 has almost the same (/proc/sysrq-trigger) ChangeSet@1.1065.1.48, 2003-08-25 16:30:09-03:00, acme@conectiva.com.br [PATCH] irqreturn_t compatibility with 2.6 Hi Marcelo, As I described to you in another message this patch is needed so that drivers don't have to be adding private copies of irqreturn_t and friends, like ATM, aic7xxx already do in 2.4.22-pre10, please apply. It is still against 2.4.22-pre10, but I think it'll apply cleanly, holler if not. - Arnaldo ChangeSet@1.1065.1.47, 2003-08-25 16:25:44-03:00, m.c.p@wolk-project.de [PATCH] Allow console switching after kernel panic() Hi Marcelo, if a kernel panic()'s, there's no chance to switch to another console for maybe important informations either about the panic or some important email the user was writing or such stuff. Attached patch allows _only_ console switching if a panic occured. Patch from Andi Kleen! It's in -wolk for a long time. ChangeSet@1.1065.1.46, 2003-08-25 15:37:31-03:00, marcelo@logos.cnet Cset exclude: m.c.p@wolk-project.de|ChangeSet|20030825183254|28555 ChangeSet@1.1065.1.45, 2003-08-25 15:32:54-03:00, m.c.p@wolk-project.de [PATCH] disable READA is an SMP only thing. Make it so. [ The following text is in the "iso-8859-15" character set. ] [ Your display is set for the "ISO-8859-1" character set. ] [ Some characters may be displayed incorrectly. ] Hi Marcelo, $subject says it all. It's in -aa, -wolk for a long time. Reference: http://lists.insecure.org/lists/linux-kernel/2002/Jul/6615.html ChangeSet@1.1065.7.4, 2003-08-25 13:19:37-05:00, shaggy@shaggy.austin.ibm.com JFS: add nointegrity mount option (Karl Rister) This option allows a performance boost by not writing to the journal at the expense of loss of data integrity if the system crashes or is powered off. The intended use of this option is to allow faster restores from backup media. ChangeSet@1.1065.1.44, 2003-08-25 15:15:54-03:00, m.c.p@wolk-project.de [PATCH] Replace bogus and obsolete "#if __SMP__" -> CONFIG_SMP Hi Marcelo, there are some places in the current kernel where __SMP__ appears. This is from old 2.2 times and is never ever defined nor used in current 2.4. Attached patch replaces __SMP__ with CONFIG_SMP. It also cleanes up 3 bogus #ifdef __SMP__ in fs/jbd code which is never used and not needed. Patch from Andrea Arcangeli! It's in -aa since 2.4.21-aa1, in -wolk for some time. Same cleanup was made to 2.5.x. Please apply for 2.4.23-pre1. Thank you :) With this applied, _all_ bogus __SMP__ defines are gone from 2.4. This patch is absolutely safe and just a cleanup! ChangeSet@1.1065.1.43, 2003-08-25 15:11:30-03:00, m.c.p@wolk-project.de [PATCH] Cleanup DRM submenu Hi Marcelo, attached patch cleans up the DRM submenu. ChangeSet@1.1065.1.42, 2003-08-25 15:04:58-03:00, m.c.p@wolk-project.de [PATCH] Fix initrd with netboot From "Martin Persson" we've encountered a problem when trying to netboot 2.4.21 and above. After /linuxrc has been executed and the kernel tries to remount the root, it panics with the all too well known message "Unable to mount root fs on ...". The kernel bugs out in mount_block_root in the file init/do_mounts.c, to be more precise in the for-loop. What happens is that it tries to mount the file system as type ext2 (which happens to be first in the list in our case), but instead of returning -EINVAL it returns -EBUSY, the loop exits instead of trying the next (correct) fs-type and the kernel panics. In -wolk for some time and maybe others. ChangeSet@1.1065.1.41, 2003-08-25 15:03:11-03:00, m.c.p@wolk-project.de [PATCH] Cleanup kmem_cache_reap() From "fbl@netbank.com.br" # This patch includes the following deltas: # ChangeSet 1.1193 -> 1.1194 # mm/slab.c 1.22 -> 1.23 # # The following is the BitKeeper ChangeSet Log # -------------------------------------------- # 03/06/03 fbl@netbank.com.br 1.1194 # [PATCH] Cleanup kmem_cache_reap() # # There is no need to keep the list_entry() # out of DEBUG condition. # -------------------------------------------- In 2.5 and -wolk for some time. # This is a BitKeeper generated patch for the following project: # Project Name: Linux kernel tree # This patch format is intended for GNU patch command version 2.5 or higher. # This patch includes the following deltas: # ChangeSet 1.1193 -> 1.1194 # mm/slab.c 1.22 -> 1.23 # # The following is the BitKeeper ChangeSet Log # -------------------------------------------- # 03/06/03 fbl@netbank.com.br 1.1194 # [PATCH] Cleanup kmem_cache_reap() # # There is no need to keep the list_entry() # out of DEBUG condition. # -------------------------------------------- # ChangeSet@1.1065.1.40, 2003-08-25 14:46:28-03:00, hch@lst.de [PATCH] fix copy_namespace() Originally by aeb, in 2.5 for ages. Fix far too small kmalloc in copy_namespace and promote errors properly. ChangeSet@1.1065.1.39, 2003-08-25 14:37:52-03:00, Jack_Hammer@adaptec.com [PATCH] ServeRAID 6.10 Driver Update With the completion of months of testing at both Adaptec and IBM, the newest version of the ServeRAID Linux device driver( ips Version 6.10.24 ) iscompleted and ready to ship. Attached is the full source ( ips.c & ips.h )and patches for the 2.4 Linux kernel. Both 32-bit andia64 architectures are supported. Please update all 2.4.x kernels with this latest driver as soon as possible. Updates for the 2.5 kernel are being handled separately. Let me know if you have any problems. Jack Hammer ChangeSet@1.1065.1.38, 2003-08-25 14:21:29-03:00, marcelo@logos.cnet pcwd.c: fix oops on unload From Arkadiusz Miskiewicz This patch is from Alan Cox and fixes problems when pcwd driver is loaded while there is no pcwd hardware installed. No idea why it was not feeded to you. Problem is described here http://www.ussg.iu.edu/hypermail/linux/kernel/0308.1/0178.html and tracked down in private conversation with Cox. ChangeSet@1.1065.1.37, 2003-08-25 14:16:14-03:00, mmagallo@debian.org [PATCH] AGPGART support for Intel 7x05 chipsets (backported from 2.6) Hi Marcelo, can you please apply the attached patch before releasing 2.4.22? It adds AGPGART support for Intel's 7205 and 7505 chipsets, which are becoming increasingly common on the street. I have checked on several machines that the patch works, even with AGP 8x. The only hardware I have access to is from NVIDIA, which means I haven't been able to test this with a non-tainted kernel, but I don't expect this to be a problem. ChangeSet@1.1065.1.36, 2003-08-25 13:58:04-03:00, willy@w.ods.org [PATCH] Fix amd67x_pm.c crash with no chipsets / CONFIG_HOTPLUG amd76x_pm.c will crash if no chipsets are found and CONFIG_HOTPLUG is turned on. This patch makes it return with a failure instead. ChangeSet@1.1065.1.35, 2003-08-25 13:57:10-03:00, marcelo@logos.cnet Changed Makefile to 2.4.22 TAG: v2.4.22