Hello Roland! This is a patch for ehca_irq.c that fixes an unproper use of spin_unlock in irq handler. Thanks Nam Signed-off-by Hoang-Nam Nguyen <hnguyen@de.ibm.com> --- ehca_irq.c | 2 +- 1 files changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c b/drivers/infiniband/hw/ehca/ehca_irq.c index e7209af..93788d8 100644 --- a/drivers/infiniband/hw/ehca/ehca_irq.c +++ b/drivers/infiniband/hw/ehca/ehca_irq.c @@ -440,7 +440,7 @@ void ...

Hello Roland! Here is patch set for ehca with the following bug fixes: * Fix unproper use of yield within spinlock context * Fix mismatched spin_unlock in irq handler Thanks Nam ehca_cq.c | 5 ++++- ehca_irq.c | 2 +- 2 files changed, 5 insertions(+), 2 deletions(-) -

at least your interrupts look sane. I see you are using MSI, but no interrupts arrive at that won't work. You either need to start with 2.6.20-rc5 (and pull the changes pending merge in netdev-2.6 from Jeff Garzik), or start with 2.6.20-rc4-mm1 and manually apply that patch I sent out on monday. A different combination of either of these two will not work, as they are completely different drivers. can you include `ethtool ethX` output of the link down message and `ethtool -d ethX` as ...

I generally tend do cleanups as a groundwork for the real fixes and separate cleanups and fixes to have good base for dealing with regressions. Often all changes (cleanups/fixes) could be included in one patch but then I would have had harsh times when debugging the regressions. It matters a lot if you hit an unknown (or known but the documentation is covered by NDA) hardware bug - you can concentrate on a small patch changing the way in which hardware is accessed instead of that big patch ...

Yes, it seems that there are actually two bugs here: * the maximum allowed PIO mode should be PIO4 not PIO5 * for auto-tuning ("pio" == 255) it incorrectly sets PIO0 It is side-effect of basing atiixp on piix driver. Fixing it will allow PIO1 to be used (good) because atiixp_dma_2_pio() always downgrades PIO1 to PIO0 Using ->tuneproc() will also set the PIO mode on the drive Yep, but it seems to be even more complicated since config_art_rwp_pio() is a mess^2 - chipset is ...

The commit 4916b3a57fc94664677d439b911b8aaf86c7ec23 introduced a hid regression between 2.6.19 and 2.6.20-rc1. The device put in input_dev->cdev is now of type usb_device instead of usb_interface. No ID_MODEL, ID_VENDOR, ID_REVISION, ID_TYPE etc etc. Fix this by assigning the intf->dev into hid->dev, and fixing all the users. Signed-off-by: Anssi Hannula <anssi.hannula@gmail.com> --- I recommend this fix to go to the stable tree before 2.6.20 is released. diff -Nurp -x ...

The reclaimer is passed a control structure that can be used to disable The traditional reclaim works fine if told what to do. Introducing another LRU list to do reclaim is a significant change to the VM, creates lots of overhead etc. -

Hi kernel developers I don't know if this is the proper list but I have a very odd problem and it's driving me nuts for the past two days. I have a portable mp3 player named: usb-storage: waiting for device to settle before scanning Vendor: THOMSON Model: PDP95FM Series Rev: 0100 Type: Direct-Access ANSI SCSI revision: 04 usb-storage: device scan complete SCSI device sda: 244288 2048-byte hdwr sectors (500 MB) sda: Write Protect is off sda: Mode Sense: 3e ...

Good idea. -

--- include/linux/cdev.h | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/include/linux/cdev.h b/include/linux/cdev.h index f309b00..b53e2a0 100644 --- a/include/linux/cdev.h +++ b/include/linux/cdev.h @@ -5,6 +5,7 @@ #include <linux/kobject.h> #include <linux/kdev_t.h> #include <linux/list.h> +#include <linux/fs.h> struct cdev { struct kobject kobj; -- 1.4.4.1 -

NAK. Better is this: Add missing struct predeclarations, otherwise we may get x.c:2: warning:

----- And on FC5 I am using pthread_self but my problem isn't simply with pthread_self, it's with the scheduling. On FC3 both threads run simultaneously in almost symmetric parallel. On FC5 one thread don't pick up and start until the previous one is done. On FC3, using getpid for the thread I could use set_afinity to force each thread to its own processor and with FC5 I can't; so I've got one idle processor all the time. -brian Brian McGrew { brian@visionpro.com || ...

maybe your kernel has a broken scheduler loadbalancing? you really again you can use gettid() or pthread_self() in that call (but remember it's a bitmask not a number); but really you shouldn't have to do this. Try a kernel which has a non-broken load balancer? -- if you want to mail me at work (you don't), use arjan (at) linux.intel.com Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org -

I do that with CF and memory stick, what kernel, distribution, etc? I would suspect something in hotplug not noticing. I presume you have tried this from cold boot, so any issues with unplugging and plugging another flash are removed. -- bill davidsen <davidsen@tmr.com> CTO TMR Associates, Inc Doing interesting things with small computers since 1979 -

Greetings all; I have a card reader plugged into a usb port. I recognizes a 512meg pny cf card just fine, but wwhen I plug in a 256meg Lexar cf, the led comes on, but there is no reaction from linux. /dev/sda is not created, nothing. Is this a kernel config problem, or is this particular cf known to be a bad bird? -- Cheers, Gene "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Yahoo.com and ...

I have been running 2.6.20-rc2-mm1 without problems, but both rc3-mm1 and rc4-mm1 have been giving me these freezes. They were happening inside X and without external console it was impossible to get anything, plus I was reluctant to test it since the freeze sometimes requires a full fsck.reiser4 --build-fs to recover the filesystem. But I finally got some output in a console session. I wasn't able to get it all, I made some notes of what I think the problem is. I may try again later once I ...

hey Jens, Just a poke since I haven't seen this change go into your block tree. Is it still in-plan? -- dann frazier -

NAK. I don't see any place in the kernel where we need a non-forced inline. We have tons of inline's in C files that should simply be removed - Oh, and your patch would have been broken for gcc < 4. 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 -

In the kernel it's what you should expect since it's defined this way It's the definition of a broken option. Not Q.E.D. due to "rightly or wrongly". It could be because Alpha uses "extern inline" and with it's old semantice I'd understand that always_inline might be a problem - but is there actually any place where Alpha uses "extern inline" with this Can you give at least one concrete example of actually broken code? The only implication I know it has caused is increased code ...

Hey Roland, Jeff has pulled in the Chelsio Ethernet driver. If you are ready to merge in the RDMA driver, you can pull it from git://staging.openfabrics.org/~swise/cxgb3.git for-roland Thanks, Steve. -------- Forwarded Message -------- From: Jeff Garzik <jeff@garzik.org> To: Divy Le Ray <divy@chelsio.com> Cc: netdev@vger.kernel.org, linux-kernel@vger.kernel.org, swise@opengridcomputing.com Subject: Re: [PATCH 1/10] cxgb3 - main header files Date: Thu, 18 Jan 2007 22:05:02 ...

Yes, from this viewpoint more reasonable approach would be to untie corresponding ptes from swap entry and mark them as invalid to trigger fault on access. Not even tried simply because it is definitely not that thing, which we needed. We used this for process migration and for that purpose we really need to know when swapoff() fails ASAP to abort migration, to kill processes which got invalid pages and to resume original copy. Obviously, delayed fault is absolutely inappropriate for this ...

So it sounds like you've been advocating a general discussion of this issue for a few years now. To summarize the issue: 1) users of the block layer assume that it's fine to associate pages that have a zero reference count with a bio before requesting I/O, 2) intermediaries like iscsi, aoe, and drbd, associate the pages with the frags of skbuffs, but 3) when the network layer has to linearize the skbuff for a network device that doesn't support scatter ...

Hello, I just tried my shiny new usb extension cable (repeater): Jan 19 16:01:17 epia kernel: usb 5-1: new high speed USB device using ehci_hcd and address 60 Jan 19 16:01:17 epia kernel: usb 5-1: configuration #1 chosen from 1 choice Jan 19 16:01:17 epia kernel: hub 5-1:1.0: USB hub found Jan 19 16:01:17 epia kernel: hub 5-1:1.0: 4 ports detected Jan 19 16:01:18 epia kernel: hub 5-1:1.0: Cannot enable port 1. Maybe the USB cable is bad? Jan 19 16:01:22 epia last message repeated 3 ...

Hi, all. On early boot stage linux maps only 8M(arch/i386/kernel/head.S). It creates identity map and kernel map like this: 0x00000000 - 0x007FFFFF -> 0x00000000 -> 0x007FFFFF /* identity map */ 0xC0000000 - 0xC07FFFFF -> 0x00000000 -> 0x007FFFFF /* kernel map */ On early boot stage identity map is needed for jumping to 0xC0000000. Intel manual says, that after such jumping identity map can be deleted. (arch/i386/mm/init.c) Linux deletes early boot identity map, but when it maps 1G(all ...

Hi, I found quickly booted into a 2.6.19-rc5 kjernel which was lying around her= e=20 and here CONFIG_USB_SUSPEND doesn't make any problems with my scanner... gunzip /proc/config.gz -c|grep USB|grep -v ...

right, and that last part explains that snippet i previously posted from include/asm-alpha/compiler.h ======================== #ifdef __KERNEL__ /* Some idiots over in <linux/compiler.h> thought inline should imply always_inline. This breaks stuff. We'll include this file whenever we run into such problems. */ ======================== which is a result of this from include/linux/compiler.h: ======================== #define inline inline ...

inline, __inline and __inline__ are equivalent as far as GCC is I suspect you're thinking of a different issue. In C99, static inline means the same as in GNU89, non-static non-extern inline means 'use this definition, that does not define objects with static storage nor references identifiers with internal linkage, for inlining or for a local definition, but make calls to it fast and don't generate any out-of-line definition', and extern inline means 'compile this code into a global ...

The kernel does not use strict C89, it uses GNUC89. Andreas. -- Andreas Schwab, SuSE Labs, schwab@suse.de SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany PGP key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5 "And now for something completely different." -

ok, so based on that and a bit more surfing, i see that either "__inline" or "__inline__" are acceptable variants in gcc, and there is no distinction between them, is that right? but in terms of strict C89 compatibility, it would seem to be a bit late for that given: $ grep -r "static inline " . no? rday -

oh, *that* part i knew. what i don't understand is the difference between "inline", "__inline" and "__inline__". you can see in include/linux/compiler-gcc4.h: #ifdef CONFIG_FORCED_INLINING # undef inline # undef __inline__ # undef __inline # define inline inline __attribute__((always_inline)) # define __inline__ __inline__ __attribute__((always_inline)) # define __inline __inline __attribute__((always_inline)) #endif ...

[RFC][PATCH] Power S3 Resume optimisation=20 Here is a simple patch for optimising the S3 resume. With this patch the resume time is 0.85. Given the fact that device initialisation on the resume takes almost 70% of time, By executing the whole "device_resume()" function on a seperate kernel thread, the resume gets completed( ie. the user can precieve) by ~0.85 sec. To avoid any possible race condition while processing the IO request and to make sure all the io request are queued till the ...

(to apply this patch, handle-cpu_lock_acquire-and-cpu_lock_release-in-workqueue_cpu_callback.patch should be dropped, it is not needed any longer). With this patch workqueue_mutex is used only to protect workqueues list, all workqueue operations can run in parallel with cpuhotplug callback path. take_over_work(), migrate_sequence, CPU_LOCK_ACQUIRE/RELEASE go away. CPU_DEAD just sets cwq->should_stop and returns. After that cwq->thread runs unbound until it flushes cwq->worklist, then exits. ...

Auke Kok [auke-jan.h.kok@intel.com] wrote: | Sukadev Bhattiprolu wrote: | >I get following panic on 2.6.20-rc4-mm1 on a 2-cpu AMD Opteron system. | > | >Same basic config file seems to work with 2.6.20-rc2-mm1 on this same | >system. Have not tried -rc3-mm1 yet. | > | >Attached are config file and "lspci -vv" output. Let me know if you need | >more info. | > | >Suka | > | >--- | > | >[ 168.925840] Freeing unused kernel memory: 320k freed | > * INIT: version 2.86 booting | > * Starting ...

Hi, I believe this is one of the bugs that is fixed in the patch that I sent monday. Please try again with the patch applied to your tree and re-test. Thanks. I didn't see Andrew merge the patch yet. see: http://lkml.org/lkml/2007/1/16/226 I can mail the patch if you can't find it. Just ping me privately. Cheers, Auke -

I think we really should have support for doing cache flushes automatically on fsync, etc. User space code should not have to worry about this problem, it's pretty silly that for example MySQL has to advise people to use hdparm -W 0 to disable the write cache on their IDE drives in order to get proper data integrity guarantees - and disabling the cache on IDE without command queueing really slaughters the performance, unnecessarily in this case. There may be some cases where the ...

[PATCH] ide: disable DMA in ->ide_dma_check for "no IORDY" case If DMA is unsupported ->ide_dma_check should disable DMA. Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> --- drivers/ide/pci/aec62xx.c | 8 +++----- drivers/ide/pci/atiixp.c | 5 ++--- drivers/ide/pci/cmd64x.c | 8 +++----- drivers/ide/pci/cs5535.c | 5 ++--- drivers/ide/pci/hpt34x.c | 8 +++----- drivers/ide/pci/hpt366.c | 8 +++----- ...

[PATCH] ide: add ide_set_dma() helper * add ide_set_dma() helper and make ide_hwif_t.ide_dma_check return -1 when DMA needs to be disabled (== need to call ->ide_dma_off_quietly) 0 when DMA needs to be enabled (== need to call ->ide_dma_on) 1 when DMA setting shouldn't be changed * fix IDE code to use ide_set_dma() instead if using ->ide_dma_check directly Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> --- drivers/ide/arm/icside.c | 5 +---- ...

[PATCH] ACPI support for IDE devices This patch implements ACPI integration for generic IDE devices. The ACPI spec mandates that some methods are called during suspend and resume. And consequently there most modern Laptops cannot resume properly without it. According to the spec, we should call '_GTM' (Get Timing) upon suspend to store the current IDE adapter settings. Upon resume we should call '_STM' (Set Timing) to initialize the adapter with the stored settings; afterwards '_GTF' (Get ...

[PATCH] ide: fix UDMA/MWDMA/SWDMA masks * use 0x00 instead of 0x80 to disable ->{ultra,mwdma,swdma}_mask * add udma_mask field to ide_pci_device_t and use it to initialize ->ultra_mask in aec62xx, pdc202xx_new and pdc202xx_old drivers * fix UDMA masks to match with chipset specific *_ratemask() (alim15x3, hpt366, serverworks and siimage drivers need UDMA mask filtering method - done in the next patch) Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> --- ...

[PATCH] ide: convert ide_hwif_t.mmio into flag All users of ->mmio == 1 are gone so convert ->mmio into flag. Noticed by Alan Cox. Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> --- drivers/ide/arm/icside.c | 2 +- drivers/ide/arm/rapide.c | 2 +- drivers/ide/cris/ide-cris.c | 2 +- drivers/ide/h8300/ide-h8300.c | 2 +- drivers/ide/ide-dma.c | 8 ++++---- drivers/ide/ide.c | 5 ++--- drivers/ide/legacy/buddha.c | 2 ...

[PATCH] ide: make ide_hwif_t.ide_dma_{host_off,off_quietly} void * since ide_hwif_t.ide_dma_{host_off,off_quietly} always return '0' make these functions void and while at it drop "ide_" prefix * fix comment for __ide_dma_off_quietly() * make __ide_dma_{host_off,off_quietly,off}() void and drop "__" prefix Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> --- drivers/ide/arm/icside.c | 10 ++++------ drivers/ide/cris/ide-cris.c | 14 ++++++-------- ...

[PATCH] sgiioc4: fix sgiioc4_ide_dma_check() to enable/disable DMA properly * use sgiioc4_ide_dma_{on,off_quietly}() instead of changing drive->using_dma directly * fix warning message * add FIXME Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> --- drivers/ide/pci/sgiioc4.c | 26 +++++++++++--------------- 1 file changed, 11 insertions(+), 15 deletions(-) Index: b/drivers/ide/pci/sgiioc4.c =================================================================== --- ...

Hi, I've just updated IDE quilt tree: http://kernel.org/pub/linux/kernel/people/bart/pata-2.6/patches/ New patches: * IDE driver for Delkin/Lexar/ASKA/Workbit/etc. CardBus CF adapters (Mark Lord <mlord@pobox.com>) * ACPI support for IDE devices (Hannes Reinecke <hare@suse.de>) * ide: unregister ide-pnp driver on unload (Tejun Heo <htejun@gmail.com>) * via82cxxx/pata_via: correct PCI_DEVICE_ID_VIA_SATA_EIDE ID and add support for CX700 and 8237S (Josepch Chan ...

[PATCH] ide: make ide_hwif_t.ide_dma_host_on void * since ide_hwif_t.ide_dma_host_on is called either when drive->using_dma == 1 or when return value is discarded make it void, also drop "ide_" prefix * make __ide_dma_host_on() void and drop "__" prefix Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> --- drivers/ide/arm/icside.c | 8 ++++---- drivers/ide/cris/ide-cris.c | 2 +- drivers/ide/ide-dma.c | 17 +++++++---------- drivers/ide/ide-iops.c ...

That would imply that disabling adma via a module parameter should make the issue go away, right? I'll try to have a test run with adma disabled over night then. Thanks, Björn -

That was my assumption as well, but was hoping someone could decode the above information and point me to the problem chip. I ran Memtest86 overnight but found no problems, but don't know if it needs to run in a particular ECC mode. This is a dual proc 275 system with 4 1GB DIMMs. Guessing that MC1 is the controller on the second CPU. Would row 1 be the second DIMM? -

Security fixes since 2.6.16.37: - CVE-2006-4814: Fix incorrect user space access locking in mincore() - CVE-2006-5173: i386: save/restore eflags in context switch - CVE-2006-5749: Call init_timer() for ISDN PPP CCP reset state timer - CVE-2006-5755: x86_64: Don't leak NT bit into next task - CVE-2006-5757/CVE-2006-6060: grow_buffers() infinite loop fix - CVE-2006-5823: corrupted cramfs filesystems cause kernel oops - CVE-2006-6053: handle ext3 directory corruption better - CVE-2006-6054: ext2: ...

While functionality is extremely interesting.... does it really have Heh. We'll waste power trying to save it. If you have to hook it into kernel, can you at least do it properly? Pavel -- (english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html -

Any idea how much power this saves? And for the vast rest of us who do run X, this seems to parallel the work of a well-tuned screensaver. -- bill davidsen <davidsen@tmr.com> CTO TMR Associates, Inc Doing interesting things with small computers since 1979 -

--- The code calls it "System Inactivity Notifier". scott -

Hi, Please use kernel CodingStyle in the kernel Documentation/ directory. That would mean /*...*/ comments instead of // and expanding the if/else lines to have their statements on separate Thanks. --- ~Randy -

I can't see any Hypervisors ever allowing those weird MSRs, so for paravirtualization it is probably better to just disable then. -Andi -

Thanks for the pointers! I've bookmarked them for future reference, and sent a new patch that follows the guidelines. :) -- Ben -

On Thu, Jan 18, 2007 at 02:13:39PM -0500, Benjamin Romer wrote: This patch looks good to me. You might want to provide some description too for changelog. Thanks Vivek -

Sslab deals with those issues the right way. We have per processor queues that attempt to keep the cache hot state. A special shared queue exists between neighboring processors to facilitate exchange of objects between then. -

Swap is usually enabled at startup, but sure, if you want you can mess All users of the NVidiot crap deserve all the pain they get. How so? I cover pretty much every allocation needed to process an skb by setting PF_MEMALLOC - the only drawback there is that the reserve might not actually be large enough because it covers more allocations that were considered. (thats one of the TODO items, validate the reserve But the network can create OOM conditions for itself just fine. Consider ...

Replace kmap() with kmap_atomic(). Reduce the amount of time that mappings are held. Signed-off-by: Michael Halcrow <mhalcrow@us.ibm.com> Signed-off-by: Trevor Highland <tshighla@us.ibm.com> --- fs/ecryptfs/crypto.c | 8 +-- fs/ecryptfs/ecryptfs_kernel.h | 4 - fs/ecryptfs/mmap.c | 121 ++++++++++++----------------------------- 3 files changed, 39 insertions(+), 94 deletions(-) ff2dcd23fe6bd000049987b2bf98ada70755f114 diff --git a/fs/ecryptfs/crypto.c ...

Hard. google(ext3 reservations) is a starting point. -

Sorry, this patch would cause build failure if setup-bus.c was not built into kernel. Revised again. Subject: [PATCH] Make CARDBUS_MEM_SIZE and CARDBUS_IO_SIZE customizable CARDBUS_MEM_SIZE was increased to 64MB on 2.6.20-rc2, but larger size might result in allocation failure for the reserving itself on some platforms (for example typical 32bit MIPS). Make it (and CARDBUS_IO_SIZE too) customizable by "pci=" option for such platforms. Signed-off-by: Atsushi Nemoto ...

... and maybe we should just kill /dev/kmem entirely... it seems mostly used by rootkits but very few other things, if any at all... -- if you want to mail me at work (you don't), use arjan (at) linux.intel.com Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org -

Ok, I'll do that, thanks! Regards, Nadia -

Hi, I've been confused by 'vma->vm_pgoff' meaning. I assumed that it was an offset relatif to the start of the kernel address space (PAGE_OFFSET) as the commit message I submitted explains. So doing: fd = open("/dev/kmem", O_RDONLY); kmem = mmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 4 * 4096); actually asks for a kernel space mapping which start 4 pages after the begining of the kernel memory space. So yes, if the 'offset' expected by 'mmap(/dev/kmem, ..., offset)' usage is ...

I agree that the name "kmem" lends itself to that interpretation (and the 2.4 through 2.6.11 history shows that's not a wild idea); but read and write of /dev/kmem have always used the kernel virtual Apology surely accepted, it's a confusing area (inevitably: in a driver for mem, the distinction between addresses and offsets become blurred). But please note, the worst of it was, that your patch comment gave no hint that you were knowingly changing its behaviour on the "main" architectures: ...

> But personally I'd prefer it to remain. Similar. I also got some tools who use it to read kernel variables and doing the V->P conversion in user space would be tedious and unreliable (e.g. for vmalloc) -Andi -

The disk seems flash disk, and this entry may be on the middle of erasing block. The flash disk might be unplugged before flushing dirty -- OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> -

This is just to enable the error reporting of the device. One of the important parts of AER is error recovery. When an AER error happens, device will send an error message to root port. Then, root port will notify kernel by interrupt. Kernel will print out error message, and do error recovery to recover the related devices. Such recovery need device drivers to provide a couple of callbacks. Documentation/pci-error-recovery.txt Usually, 3 callbacks are enough while error_detected is must. ...

Hi I found the reason for the long disabling of Soft-Irqs: In-between the spin_lock_bh() and spin_unlock_bh() was a printk() going to serial console with 19200baud. So it took easily 30mS for console-output. No i start klogd an the log Messages go to a file in tmpfs in a short time. Juergen Pfeiffer, Seskion GmbH Karlsruher Str. 11/1 70771 ...

Hi Aubrey, I used your patch on my PPC64 box and I do not get expected behavior. As you had requested, I am attaching zoneinfo and meminfo dumps: # cat /proc/sys/vm/pagecache_ratio 50 # cat /proc/meminfo MemTotal: 1014600 kB << 1GB Ram MemFree: 960336 kB << Expect to see around 500MB free after Buffers: 8348 kB issue of DD command Cached: 8624 kB SwapCached: 8 kB Active: 20908 kB Inactive: 5680 kB SwapTotal: ...

Hi Aubrey, The idea of creating separate flag for pagecache in page_alloc is interesting. The good part is that you flag watermark low and the zone reclaimer will do the rest of the job. However when the zone reclaimer starts to reclaim pages, it will remove all cold pages and not specifically pagecache pages. This may affect performance of applications. One possible solution to this reclaim is to use scan control fields and ask the shrink_page_list() and shrink_active_list() ...

Although I have no PPC64 box in hand, I think the logic should be the same. get_page_from_freelist() is called 5 times in __alloc_pages(). 1) alloc_flags = ALLOC_WMARK_LOW | ALLOC_PAGECACHE; 2) alloc_flags = ALLOC_WMARK_MIN | ALLOC_PAGECACHE; We should have the same result on the first two times get_page_from_freelist(). 3) if (((p->flags & PF_MEMALLOC) || unlikely(test_thread_flag(TIF_MEMDIE))) && !in_interrupt()) alloc_flags = ALLOC_NO_WATERMARKS The case on my platform will never ...

That would be good as a default, but I've been thinking that we could perhaps also add a sysctl in /proc/sys/fs/nfs in order to make it a tunable? Cheers, Trond -

Eeuh, right. Glad to have you around to remind how puny my boxens Right, perhaps cap this to a max of 256M. That would allow 128 2M RPC transfers, much more would not be needed I guess. Trond? -

balance_dirty_pages() wbc.sync_mode = WB_SYNC_NONE; writeback_inodes() sync_sb_inodes() __writeback_single_inode() __sync_single_inode() write_inode() nfs_write_inode() Ah, yes, I see. That ought to work. /me goes verify he didn't mess it up himself... -

Ah, yeah, forgot to refresh the patch one last time before sending out :-(. -

No, socket locking is not solution at all here. But the same applies to header - it will be copied into socket queue, then socket will be unlocked and populated VFS data will be put into that queue too, but there is a window between socket unlock after header copy and file data copy. If we will hold socket lock after header is copied, it is possible to lock it for too long - bad sectors on disk, I bet those people worked with different Unix systems, which have much slower syscalls, so they ...

Hi Dann ! Thanks ! Best regards, Willy -

printk("ACPI: retry %d\n") -> printk("ACPI: retry %d\n", retry) ;) ill try this again soon. -

Ok, here is what i got: hwacpi-0207 [C031D380] [04] hw_get_mode : ----Entry hwregs-0273 [C031D380] [05] get_register : ----Entry hwregs-0487 [C031D380] [06] hw_register_read : ----Entry hwregs-0810 [C031D380] [06] hw_low_level_read : Read: 00000000 width 16 from 0000000000000404 (SystemIO) hwregs-0575 [C031D380] [06] hw_register_read : ----Exit- AE_OK hwregs-0300 [C031D380] [05] get_register : Read value 00000000 register 3 ...

Yes, that is one of the nice things about 64-bit machines. No special cases - you sure can get 30GB if your machine is equipped with that much. Or you can get that much virtual memory if you have enough swap . . . Helge Hafting -

Not for me. I'm still have the same trouble, but less (maybe about every hour, instead of every 5 minutes). futhermore, I found a patch cocktail-2.6.20-rc3.patch: http://tinyurl.com/2gza8q, which improves the situation too! Now, the funny thing is that I've two SATA HDDs, but only 1 causes all the headaches. The affected drive is a: sda - @ata3.0 - WDC WD2500KS-00M 02.0 ATA-7, max UDMA/133, 488395055 sectors: LBA48 "ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 ...

I have exactly the same problem on -rc5 and it causes all I/O to stall periodically if I do _anything_ I/O intensive. On my box, I have 4 sata_nv handled SATA ports, with two pairs of different drives (two Maxtor, two WD) and it happens randomly on both. So it's absolutely nothing to do with the drive make/model. I'll try Jeff's suggestion of disabling ADMA now, but I think something more radical than this workaround should make it into 2.6.20 final, otherwise a lot of people are going ...

I think the exact placement of these files needs a bit of work. I don't much care about xen/ vs paravirt/xen vs paravirt-xen/, but placement of headers needs a bit more thought. lhype, er, lguest and vmi are in kernel/, but I think it would be nice to put them near Xen. J -

Linus, please pull from git://git.infradead.org/mtd-2.6.git This fixes the above bug along with a few others. It does also contain a small amount of new code which has been waiting for a while (including the driver for the CAFÉ NAND controller which we use on OLPC.). My apologies for missing the merge window and first asking you to pull this a few hours after 2.6.20-rc1 was cut; I'd been waiting for the bitrev stuff to land, and had waited too long. Adrian Bunk (3): [MTD] SSFDC ...

This patch contains the following possible cleanups: - make needlessly global code static - remove the following unused variable: - debug.c: alloc_prints Signed-off-by: Adrian Bunk <bunk@stusta.de> --- BTW: Do you really need that many different CONFIG_MTD_UBI_DEBUG_* options? drivers/mtd/ubi/cdev.c | 4 +- drivers/mtd/ubi/debug.c | 57 ++++++++++++++++++---------------------- 2 files changed, 28 insertions(+), 33 deletions(-) --- ...

Hello. BTW, if I'm (or somebody else) sending patches with IDE driver *fixes*, should these be against that quilt tree (containing largely reworked IDE code), against fresh Linus' tree (or some recent -rc), or against recent -mm tree? WBR, Sergei -

Hi Andrew, Thanks again for finding the fix to the problem I reported. Can you tell me when I might expect this fix to show up in 2.6.20-rc? Thanks, Mike -

--Boundary-01=_qnMsFdijl38wsgH Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Here you go. =2D-=20 (=C2=B0=3D =3D=C2=B0) //\ Prakash Punnoor /\\ V_/ \_V --Boundary-01=_qnMsFdijl38wsgH Content-Type: application/x-bzip2; name="lsusb.txt.bz2" Content-Transfer-Encoding: base64 Content-Disposition: ...

PCI posting (need to read a register before delaying to guarantee flush; standard technique is to use random bytes, not a fixed address like this, when MAC address is otherwise unavailable. consider what happens -

This doesn't work for me in various nasty ways. Please, check for a minimum delay or loop to get ahead of time. See code in the "old" timer ISR. See how it unconditionally adds at least 10 ticks... Thanks Guennadi --- Guennadi Liakhovetski -

For example, consider the following case: a process forks many children and system administrator wants to allow only one of these processes to dump shared memory. This is accomplished as follows: $ echo 1 > /proc/self/coremask $ ./some_program (fork children) $ echo 0 > /proc/<a child's pid>/coremask With the /proc/<pid>/ interface, we don't need to modify the user program. In contrast, with the ulimit or setrlimit interface, the administrator can't do it without modifying the ...

applied -

Greetings. It appears to me that under certain circumstances mmc layer will issue requests to the host after mmc_host_remove returns. This happens, for example, in tifm_sd driver because mmc_host may be removed mid-transfer, as the socket shall be freed for possible reuse by different media type. Currently, the only solution is to sleep a little somewhere after mmc_remove_host but before mmc_free_host. I think the correct way to handle the situation is to ensure that mmc_host is never accessed ...