Summary of changes from v2.5.19 to v2.5.20 ============================================ add pSeries hypervisor console pSeries HVC console: add hvc initalisation pSeries HVC console: Disable interrupts in spinlock regions, fix from Milton Miller. The backtrace he found was: hvc_write... tqueue_bh... do_softirq... timer_interrupt... hvc_write pSeries HVC console: add missing function prototype ppc64 signal cleanup: remove old debug and sync with ppc32 signal code ppc64: signal formatting cleanups ppc64: SIGURG fix from Chris Yeoh ppc64: cleanup 32 bit signal code ppc64: SIGURG fix from Chris Yeoh ppc64: update config.in ppc64: fixes for 2.5.14 ppc64: fix compile error caused by machine type changes ppc64: Add PREEMPT_ACTIVE and fix HMT macro typo ppc64: provide default_idle, we shouldnt ever actually use this but we need it to link. ppc64: missed this during Naca -> naca merge ppc64: dont allow tlbiel on large pages. add power4 flag for assembler, needed for tlbiel. ppc64: Add TIOCSTOP translation add subsys_initcall for pcibios_init ppc64: remove old code for stacking signals, its not used any more ppc64: reduce differences to the ppc32 signal code. ppc64: Be sure to zero the fpscr in a signal handler, otherwise we take a SIGFPE as soon as the signal handler tries to do any FP. ppc64: Fix a number of problems with our exception reporting, in particular ensure siginfo is correctly filled out. ppc64: Take page_table_lock in flush_tlb_mm ppc64: use common die() function in bad_page_fault, from ppc32 ppc64: prevsp is not used any more remove fpscr_pad, not required on 64 bit ppc64: Add Ingo's irq affinity stuff, from x86 ppc64: use generic debugger hooks instead of hardwiring xmon everywhere ppc64: create CONFIG_XMON_DEFAULT so we can have debugging enabled but xmon disabled at boot. ppc64: Paul Mackerras' optimized power4 copy routines ppc64: dont print "xmon called", it garbles the output when multiple cpus hit xmon at the same time ppc64: Up the gcc inline limited, needed for gcc 3.1 ppc64: remove old copy_page, missed during pervious merge ppc64: No need to handle address space destruction in flush_tlb_mm, its done in zap_page_range - From Paul Mackerras. ppc64: Add get/setaffinity syscalls, from sparc64 ppc64: Provide a rough cache_decay_ticks. ppc64 segment table rework. We preload segments for both segment table and SLB based machines. We support both preloading all 16 segments in 32b or just the PC and SP segments (32b and 64b). ppc64: Remove last_syscall, we can work the syscall out easily from our symbolic backtrace. ppc64 defconfig update ppc64: missed during segment handler rework Janitor: Add __devinit markers to two net drivers, epic100 and sundance [ARM] Remove redundant instruction cache code from decompressor. Now that we cleanly support handling of the cache for various ARM processors in head.S, the per-processor class include file, and associated inline function is no longer required. This changeset removes these files and the function call. [ARM] 2.5.18 Update ARM for TLB shootdown changes We basically do the same as Sparc64. [ARM] Tidy up context switch save area initialisation for fork Fix a minor bug where we jump to ret_from_fork with IRQs enabled. [ARM] Remove victor machine type This machine type is no longer maintained and is falling into bitrot heaven. [ARM] Fix makefiles for drivers/acorn/{block,char}/ [ARM] cpufreq_init takes low and high frequency limits. [ARM] Context switch improvements - We don't need to save the CPSR. - Rearrange thread_info members so we can pull the fields out of thread_info more efficiently. - Allocate a couple of extra words for CPU specific context saving (eg, for Xscale ACC registers) - Always leave 8 bytes free at the top of the kernel stack. This prevents the stack becoming completely empty when do_exit() is called from an exiting nfsd() thread, and causing the wrong pointer to be returned from current_thread_info() [ARM] Remove an extraneous load from atomic ops atomic.h was generating some extra loads that aren't required. In addition, it was needlessly performing various tests inside the atomic region. This change fixes both these issues. Device Model: Add helpers bus_for_each_dev and bus_for_each_drv - iterators for accessing bus's lists of devices and drivers - does locking on bus for each list access - does reference counting on objects as it grabs and releases them - calls callback for each object in list Device Model: Implement centralized device/driver binding - on device registration, all drivers of bus are iterated over - bus's bind callback is called to match device to driver - if successful, driver's probe callback is called - on device removal, driver's remove callback is called - on driver registration, list of devices is iterated over (and same thing happens) PCI Update: - Change pci_match_device to pci_bus_bind, which becomes bind callback of pci_bus_type - Remove explicit driver binding; let the device core handle it PCI: Make sure id_table is passed to probe callback JFS: cleanup dbAlloc Reduces number of return paths to two. This is a preparation for quota support. Submitted by Christoph Hellwig JFS: metapage changes This patch is a first preparation of using the bdev mapping instead of the JFS-private device mapping/inode for 2.4.1x/2.5: * add metapages to ->mp_list of the inode passed to __get_metapage instead of mp->mapping->host. Do this outside of add_to_hash, and only do it for metapages that aren't mapped absolutely. RATIONALE: when using the bdev mapping, the absoloute inode won't have the JFS-private inode info. It is save because we never traverse the mp_list of the direct inodes. * use a goto to unify two error handling pathes in __get_metapage. Submitted by Christoph Hellwig JFS: misc stuff for 2.5 o make JFS_SBI an inline instead of a macros, as the other filesystems do o ditto for isReadOnly o move initialization of sb into the common part of lmLogOpen Submitted by Christoph Hellwig JFS: support for kNFSD Add support for the get_parent export operation to make nfs exporting work on 2.5. Switch to the new iget_locked and eliminate ->read_inode. Submitted by Christoph Hellwig, and Dave Kleikamp Fix free-space leak truncating files in JFS. There was a bug truncating large files to non-zero lengths which led to space not being marked free in the block map. rename ACPI files to remove acpi_ prefix. Why did we ever name them that way? I don't know but it was dumb, so put them back. more file renames [PATCH] suspend-to-ram: clean up according to Andy Andy did not quite like the comments, please apply this. Clean up code based on things flagged by lint Improved error checking on hw accesses Moved arch specific macros to arch/asm/acpi.h [PATCH] Fix RAMDISK config problem The following fixes compilation with CONFIG_BLK_DEV_RAM=y I assume that Rusty intended to use a test for CONFIG_BLK_DEV_RAM and not BLOCK_DEV_RAM. [PATCH] de4x5 does not compile in 2.5.19 due to bus_type conflict de4x5.c used struct bus_type for its internal bookkeeping. It should be proably converted to struct device, but as I prefer Jeff's tulip driver, and I have no knowledge about EISA support, I only changed the structure name to make thing compile. [ARM] Convert for() delay loops to udelay() ACPI PCI IRQ Improvements: We now can boot without MPS and PIRQ! ACPI finds IO APICs, Local apics (CPUS), interrupt source overrides, the works. (Paul Diefenbaugh) Mucho thanks to Dominik Brodowski and the whole acpi-devel gang. (Note, I think this conflicts slightly with Pavel's code in i386/kernel/acpi.c but I'd like to get this applied, and then I'm working with Pavel to hopefully get the ACPI and swsusp stuff working properly.) [PATCH] drivers/net/wan/dscc4.c - gross overflow bug: array of u32 is accessed with offset in u8 units. [PATCH] Missing include in drivers/base/bus.c and drivers/pci/pci-driver.c drivers/base/bus.c and drivers/pci/pci-driver.c both have functions which are marked with __init, but didn't include directly. The following fixes that (and allows 2.5.19 to compile on PPC32). Properly (?) handle the multiple people who can find PCI root bridges ACPI Ancillary files update move arch-dependent macros from drivers/acpi/include/platform/acenv.h (where they clearly didn't belong) into arch/asm/acpi.h Code cleanups 1) Eliminate everywhere "if (0 == result)" is used, replace with "if (!result)" and the like. 2) Move headers mysteriously included in the middle of the file to the top. [ARM] Remove hard coded per-architecture memory, ramdisk and initrd We have a fair number of command line options which can be used to pass information about the target to the kernel. We also provide a default command line which the kernel will use in absence of one provided by the boot loader. We pass the hard coded options via the command line now. USB OHCI driver: Added SA1111 support Added the patch from Russell King that splits the usb-ohci driver into two pieces, enabling the sa1111 hardware to work with the driver. I also added some changes to get the usb-ohci-pci.o module to build and run properly. PCI: Put pci_match_device back for the people that are still using it. Fix up ACPI makefile that got broken by the merge. Add ETHTOOL_PHYS_ID ethtool command to linux/ethtool.h. e100 net driver update: * Added Documentation/networking/e100.txt. * Added drivers/net/Config.help text for CONFIG_E100. * Added ethtool ETHTOOL_PHYS_ID support * Made ethtool WoL advertisement more selective on h/w controller version. e1000 net driver update 1/4: * Adds support for ETHTOOL_GREGS to dump register set to ethtool. * Adds support for ETHTOOL_LEDBLINK to blink LED's on adapter from ethtool. * Bug fix: error message added for corrupted EEPROM detection. * Bug fix: when setting up link on copper interface, explicitly turn off force speed and duplex before auto-negotiation. e1000 net driver updates 2/4: * Bug fix: detect-for-hang mechanism was getting fooled by false-positives: fixed. * Removed unnecessary tx_lock from driver, decreasing CPU utilization under heavy UDP load. e1000 net driver update 3/4: * Added four new device IDs: 82545EM_COPPER, 82545EM_FIBER, 82546EB_COPPER, 82546EB_FIBER. * Added a couple magic number definitions for DMA addressable bits. * Added Power Management Suspend/Resume support. * Added reboot notifier registration to setup WOL when shutting down. e1000 net driver update 4/4: * Documentation/networking/e1000.txt: Updated with new device IDs, updated parameters, updated known issues section. * drivers/net/Config.help: Updated with new device IDs. * Added license for Makefile * Bug fix: ethtool GEEPROM was using word-addressing rather than byte-addressing. * Bug fix: Maximum Memory Read Byte Count in PCI-X config space set to minimum of what device advertises, and what the BIOS assigns, rather than the maximum, which is the device's default mode, and is incorrect and will cause problems for large (jumbo frame) read operations. Renames struct bus_type to struct de4x5_bus_type in de4x5 net driver, to avoid collision with the struct bus_type in device.h Fix pcnet32 net driver workaround for xSeries250. eepro100 net driver trivial cleanup: Extra prototype does not hurt; ifdefs do. dl2k gige net driver updates: * fix carrier and rx frame error miscounts * fix vlan bug * h/w dma bug workaround * ISR chip-based timer scheme [PATCH] USB storage: Dead code, more abort cleanups, and detached device fix Here's a patch which does three things: (1) It removes some dead code. (2) It moves more code to using the new state machine (3) It fixes a long standing bug relating to the sense data for a detached device. driverfs update: - Suggestion from Andrew Morton: use set_page_dirty instead of SetPageDirty and move to commit_write, since it's not dirty yet in prepare_write - Make we dput on file and directory removal to counter the lookup_hash when we create them USB: define usb_bus_type and register on startup USB: - Set interface's bus type as they're registered with core - Set bus_id to have device number as well as interface number USB: Move URB request code from usb.c to urb.c USB: Move synchronous message passing code from usb.c to message.c USB: Move configuration parsing code from usb.c to config.c Fix PCI irq routing to always look up the bridge swizzling if a direct lookup didn't work. driverfs: Remove default 'status' file: it had no useful read information, the commands it supported were minimal and probably broken and the comments were wrong. device model: remove flags parameter from struct device_driver::remove and fix all users devicemode: Implement driver_for_each_dev - iterate over all devices a driver has, with proper locking on driver and refcounting on devices device model: Use driver_for_each_dev to unbind drivers from its devices (now that it's implemented) Make qla1280 driver print out irq information along with bus location. ppc64: Recent firmware removes the compatible property on pci bridges. ppc64: Fix 32 bit execve to mirror recent generic changes ppc64: kill MAP_NR, dont mark free_initmem as __init ppc64: mask top 4 bytes of si_code remove bogus panic in ppc32_select ppc64: quota updates ppc64: signal32 updates from Stephen Rothwell - comment changes ppc64: more signal32 updates from Stephen Rothwell ppc64: more signal32 updates from Stephen Rothwell - replace some awful code with loops ppc64: last of the signal32 updates from Stephen Rothwell. Fantastic stuff. ppc64: Add CONFIG_CMDLINE, from ppc32 ppc64: Fix clear_user, from ppc32 [PATCH] 2.5.19 IDE 76 - Since Bartek has released finally an entierly user space based setup monitoring utility, we can finally remove the nonfunctional PROC code from the host chip drivers. We have preserved it thus far only for documentation purposes. - Use generic bus master DMA setup code. There is nothing wrong with it. - Make the ide-scsi code actually just allow for one device id per scsi host, since we are registering a host per device right now. This prevents the repetitive device recognition. Well registering an SCSI host of every single disk out there isn't the proper thing to do. I will deal with that later after my visual perception recovers again from already looking too long at the SCSI code :-). - Deal properly with host specific data mapping. (Could be that solves some problems with the driver, which where in reality kernel data corruptions.) - Homogenize flag handling for ATAPI drivers. [PATCH] APM patch for idle_period handling This patch from Laurent Latil fixes a copy and paste error. [PATCH] missing bit from signal patches The following should allow the affected architectures to build in 2.5.19 as currently there will be two definitions of copy_siginfo_to_user and if an architecture defines its own siginfo_t it MUST define copy_siginfo(). [PATCH] i386 mmx copying bug. Odd, this is the missing half of the recent fix where we prefetched too far. Not sure how this bit got dropped. Without it, we still prefetch past the end of a range of memory. Also small compiler hint microoptimisations [PATCH] dumb cpqarray init microoptimisation. [PATCH] region handling cleanups for gscd driver The usual check_region -> request_region work, plus freeing of resources on failure path. [PATCH] i386 smp tweaks. o mention a quirk in the athlon smp detection o mark some use-once paths as __init [PATCH] ps2esdi resource cleanups checks return codes, and sanitises resource allocation. [PATCH] recognise 2 extra devices for xd.c Wow, someone found some new devices for this beast.. [PATCH] epca janitor work o check request_region result o Handle failure gracefully [PATCH] h8 janitor work o Check request_region result [PATCH] check misc_register result in efirtc [PATCH] ib700wdt janitor work o Check misc_register return code o request_region() return code checking o graceful failure [PATCH] dltk driver check_region -> request_region [PATCH] mixcomwd janitor work simple region handling cleanups. [PATCH] aztech CD driver janitor work o printk levels o check_region -> request_region cleanups o allocated region releasing on failure [PATCH] esp janitor work o check request_region result o handle failure gracefully [PATCH] elsa check_region() -> request_region() [PATCH] istallion janitor work region handling cleanups [PATCH] better sizing of queue_nr_requests Patch from 2.4 sets the maximum size of the queue in a more sensible manner [PATCH] forward ioctls on raw devices to underlying devices Should allow for things like adjusting readahead on raw devices. [PATCH] rtc max_user_freq sysctl Allow user to define the upper limit of number of interrupts generated [PATCH] ite_gpio region handling cleanups [PATCH] isicom check_region() -> request_region() [PATCH] niccy region handling cleanups [PATCH] Some includes that aren't needed. Stale cruft, and what looks like a merge error. [PATCH] missing binfmt check As per the old comment, this patch adds checking of the return value of setup_arg_pages() and aborts cleanly. From 2.4 (also a missing GPL tag) [PATCH] Deliver SIGIO to FIFO and pipe devices Originally from: Jeremy Elson [PATCH] Fix deadlock in nbd Variant of a patch in 2.4 by Steven Whitehouse. [PATCH] teles3 region handling cleanups [PATCH] Conversion of uidhash to use list_t Conversion done by William Irwin a good while ago. [PATCH] hd.c compilation fix. TIMEOUT_VALUE was killed from blk.h and moved to drivers that need it (like this one) DEVICE_REQUEST is also dead, so we init the queue with the driver request handler [PATCH] gazel region handling cleanups [PATCH] egcs no longer supported. Andrew Morton found that egcs was miscompiling the per-cpu area code on x86. This patch updates the check, and documentation accordingly. [PATCH] avm_a1 check_region -> request_region cleanup [PATCH] fill in empty hisax_fcpcipnp debug statement [PATCH] struct super_block cleanup - adfs Another one from Brian Gerst. Still applies to 2.5.18 Seperates adfs_sb_info from struct super_block. [PATCH] add spinlocking to w83877f_wdt driver [PATCH] missing GPL tags [PATCH] Fix up agpgart. Don't know what exactly happened, but somehow, the ZX1 support is in your tree twice. This patch drops one of them, and also adds in Intel i845 support to make up for it 8-) [PATCH] 2.5.19 : drivers/mtd/nftlcore.c The following patch fixes a few compiler warnings, as well as provides blk_init_queue() with the appropriate arguments. Please review for inclusion. Thanks to Jens Axboe for fixing the casting hack in the previous version of this patch. Felipe W Damasio: add KERN_* to 3c501 driver printk's [PATCH] trivial keyboard driver patch This makes it possible to run the ia64 kernel both on platforms with the legacy keyboard controller and those without (as is the case for the hp zx1 platform, for example). The new code is inside #ifdef CONFIG_IA64, though it would probably be a good idea to enable it unconditionally. The patch is by Alex Williamson. [PATCH] time-offset patch On ia64 MP machines, we use the cycle counter register of each CPU to obtain fine-grained time-stamps. At boot-time, we synchronize the counters as close as possible (similar to x86, though with a different algorithm). But even with this synchronization, there is still a small (really: tiny) chance that a process bouncing from one CPU to another could observe time going backwards. To guard against this, I maintain a global variable called "last_time_offset" which keeps track of the largest time-interpolation value returned so far. Most of this is in platform-specific code (arch/ia64/kernel/time.c), but there are a handful of places in platform-independent code where this variable needs to be cleared to zero. This is what the patch below does. I didn't put it inside CONFIG_IA64 because I think this can be useful for other platforms, too. I suppose I could put it inside CONFIG_SMP though this would make the code uglier. If you think it's OK, please apply, otherwise, I'd appreciate your feedback. [PATCH] pass "page" pointer to clear_user_page()/copy_user_page() Hi Linus, Are you willing to change the interfaces of clear_user_page() and copy_user_page() so that they can receive the relevant page pointer as a separate argument? I need this on ia64 to implement the lazy-cache flushing scheme. I believe PPC would also benefit from this. --david [PATCH] softirq.c per_cpu fix GCC3.1 apparently gets confused about uninitialized sections [PATCH] agp support for i460 and zx1 cleanup The patch below adds first round of AGP support for the Intel 460 chipset This won't actually build at the moment, but I think you prefer to merge things piecemeal, and this portion of the patch is almost guaranteed to be safe (affects only ia64). airo wireless net driver updates: * support latest wireless extensions * update locking to use semaphores [PATCH] USB kernel-api documentation fix updated the kernel-api documentation USB files due to file reorg. [PATCH] ehci remove warning if no CONFIG_USB_DEBUG I just noticed a debug message will generate a needless warning when debugging is disabled. [PATCH] PATCH: USB Scanner Driver 0.4.8 and new maintainer Here's my last and final patch to the maintainer of USB Scanner Driver. Brian Beattie is now going to assume this role (thanks Brian!). Brian brings some kernel level programming so I'm sure he'll be able to conttribute to this list w/o any problems. I want to thank you all for your support and help. A couple of you sent some personal msgs regarding my departure - thank you. * 0.4.8 5/30/2002 * - Added Mustek BearPaw 2400 TA. Thanks to Sergey * Vlasov . * - Added Mustek 1200UB Plus and Mustek BearPaw 1200 CU ID's. These use * the Grandtech GT-6801 chip. Thanks to Henning * Meier-Geinitz . * - Increased Epson timeout to 60 secs as requested from * Karl Heinz Kremer . * - Changed maintainership from David E. Nelson to Brian * Beattie . Fix missing piece of uidhash list conversion Simplify tlb_flush_mmu() for exit case: makes it easier on the ia64 folks, and the ARM people didn't mind. [PATCH] 2.5.19 IDE 77 - Get rid of SELECT_DRIVE macro. Start to move all direct hardware access functions in to one place. - Get rid of SELECT_MASK macro. Realize that the mask is always equal 0. Simplify the maskproc therefore. - Get rid of GET_STAT and OK_STAT macros as well. - hpt366 cleanups by Andrej Panin. - Artop driver update by Franz Sirl. kbuild: Don't rebuild if vmlinux if nothing changed - fix We would rebuild vmlinux if the command line to do so changed since the last build. Unfortunately, we forgot to read the old command line, so we always rebuilt it. kbuild: Add (internal) prepare target Be explicit about what we need to do before we can start the recursive build, it's now listed in the "prepare" rule. kbuild: Use the real instead of a phony target if we have one o A rule which makes the include/asm symlink should have the target "include/asm", not "symlinks". Likewise for ".hdepend". (include/linux/modversions.h yet to follow) Apart from being cleaner, this additionally gives us protection when a command fails or the user interrupts execution: E.g. for the case of "make dep", if the user interrupts at some point, make will delete the target (now ".hdepend") for us, so that at the next run, make knows that it needs to make the dependencies again, while otherwise it would just use the half-finished file. o We don't need the include/asm symlink for the configuration targets, we only need it before generating dependencies. (The setup of having /usr/include/{linux,asm} symlinked to your current kernel source is already broken in 2.5, so this change doesn't break anything which wasn't broken before.) o Add .hdepend to "prepare", i.e. what needs to be done before we can start the recursive build. That means we can get rid of the gross (and not always working) hack which made "make" run "make dep" when .hdepend didn't exist yet. o Mark the *config targets phony and improve some comments. kbuild: clean up generation of modversions.h There's no good reason why we would generate include/linux/modversions.h from the top-level Makefile when CONFIG_MODVERSION=y and from Rules.make otherwise. Nor is there a good reason to call the target to do so "update-modverfile" - "include/linux/modversions.h" makes much more sense. kbuild: Clarify the CONFIG_MODVERSIONS logic Observe that defined(MODVERSIONS) == defined(CONFIG_MODVERSIONS) && defined(MODULE) and from there I step by step simplified the logic in include/linux/module.h - staying logically equivalent, but it is much more understandable now, IMO. Still added a huge comment trying to help other people understand what kind of magic happens here. kbuild: Get rid of -DMODVERSIONS, further cleanup -DMODVERSIONS isn't used anymore, so it can go. Also, after cleaning up include/linux/module.h, it's obvious that we don't need include/linux/modversions.h at all if CONFIG_MODVERSIONS not set, no need to generate it. Rules.make explicitly lists files which depend on modversions.h, since make dep cannot know about the "-include include/linux/modversions.h" which gets added to the command line. Now that we understand when it is needed, we can even get that list right ;-) Oh well, nice theory. .hdepend will touch module.h when modversions.h changes, so we still get unnecessary recompiles. We really need to switch to the new way of dependency generation, it gets all that right without even thinking about these special cases. We don't track dependencies for .ver files. In fact, we relied on that checksum would only change if the corresponding exporting C source changes. That's not true, of course, all of the included headers have say as well. So we better force the hash to be checked unconditionally every time "make dep" is run. kbuild: Group targets which need / don't need .config The main Makefile includes .config - on the other hand, it also generates .config when using make *config. This leads to recursion problems when we make the build depend on .config. (.config is generated by *config. So after running make *config, make notices one of its includes has been updated -> it restarts to read the new contents. Unfortunately, the restart runs *config again, which updates .config again. You get the picture) Therefore, we're aiming at a two phase system: If no .config exists, allow only make *config and make clean/mrproper. Afterwards, allow all targets. (It's not much different from what we have today) [PATCH] missing argument to clear_user_page in v4l David Mosberger added argument page to clear_user_page, but apparently did not scan whole tree to find callers. Please apply. It's the only such call in the non-arch specific portion of tree. Simplify uidhash table allocation - no need to make it dynamic, as it isn't even all that big. [PATCH] 2.5.19 IDE 78 - Move ide_fixstring() from ide.c to probe.c, since this is the place, where it's most used. - Remove GET_STAT() - it's not used any longer. - Remove last parameter of ide_error. Rename it to ata_error(). - Don't use ide_fixstring in qd65xx.c host chip driver. The model name is already fixed in probe.c. - Invent ata_irq_enable() for the handling of the trice nIEN bit of the control register. Consistently use ch->intrproc method every time we toggle this bit. This simply wasn't the case before! - Disable interrupts on a previous channel only when we share them indeed. - Eliminate simple drive command handling function drive_cmd. - Simplify the ioctl handler. Move it to ioctl, since that's the only place where it's actually used. [PATCH] [PATCH} 2.5.19 IDE 79 - Fix typo in sparc_v9 code, in ns87415, just introduced. - Eliminate unnecessary struct hd_drive_hob_hdr those are in reality precisely the same registers as usual. - Eliminate control_t, nowhere used type. - Unfold ide_init_drive_cmd() at the places where it's used. This makes obvious that REQ_DRIVE_CMD gets only used on the ioctl command path. [PATCH] 2.5.19 IDE 80 - Sanitize the handling of the ioctl's and fix a bug on the way in dealing with the WIN_SMART command where arguments where exchanged. - Finally sanitize ioctl further until it turned out that we could get rid of the special request type REQ_DRIVE_CMD entierly. We are now using consistently REQ_DRIVE_ACB. One hidden code path less again! - Realize the ide_end_drive_cmd can be on the REQ_DRIVE_ACB only for ioctl() to a disk. Eliminate it's usage from device type driver modules. - Remove command member from struct hd_drive_task_hdr and place it in strcut ata_taskfile. It is not common between the normal register file and HOB. We will have to introduce some helper functions for particular command types. [PATCH] 2.5.19 IDE 81 - Don't use ata_taskfiles cmd field for drive status reporting, we can now simply use drive->status instead. - Unify command type parser entries which could be unified due to the unification of corresponding interrupt handlers. - Eliminate reading parameter from ata_do_udma(). We have this information already in the rq. This allows us to merge several methods. - Rename XXX_udma to udma_setup, since we have finally settled up on this semantics. - Simplify tons of host chip code by removing wrapper functions. [PATCH] 2.5.19 IDE 82 - PPC compilation fix by Paul Mackerras. - Various fixes by Bartek: fix ata_irq_enable() and ata_reset() for legacy ATA-1 devices in start_request() for REQ_DRIVE_ACB a) don't run ->prehandler() twice b) return ata_taskfile() value [PATCH] documentation and tq_disk removals This should be the last of tq_disk, at least the trivial ones. md still has some queue_task references, I'll let Ingo/Neil clean those up. suspend is still broken, it was broken before too though. I guess Pavel will want to fix that. Also, I've documented the plug functions. [PATCH] misc generic block tag fixes A buglet and a few adjustments. [PATCH] update to the update Too much copy'n paste between 2.4 and 2.5 code base, attached patch on top of the previous block tag fixes makes it work/compile again. Sorry about that. [PATCH] Re: ufs compile error in 2.5.19] I missed this one in the last patch I sent to you. [PATCH] list_head debugging A common and very subtle bug is to use list_heads which aren't on any lists. It causes kernel memory corruption which is observed long after the offending code has executed. The patch nulls out the dangling pointers so we get a nice oops at the site of the buggy code. [PATCH] remove PageSkip() macros Remove some unused PageSkip() macros. Presumably leftovers from PG_skip which isn't there any more. [PATCH] fix swapcache packing in the radix tree First some terminology: this patch introduces a kernel-wide `pgoff_t' type. It is the index of a page into the pagecache. The thing at page->index. For most mappings it is also the offset of the page into that mapping. This type has a very distinct function in the kernel and it needs a name. I don't have any particular plans to go and migrate everything so we can support 64-bit pagecache indices on x86, but this would be the way to do it. This patch improves the packing density of swapcache pages in the radix tree. A swapcache page is identified by the `swap type' (indexes the swap device) and the `offset' (into that swap device). These two numbers are encoded into a `swp_entry_t' machine word in arch-specific code because the resulting number is placed into pagetables in a form which will generate a fault. The kernel also need to generate a pgoff_t for that page to index it into the swapper_space radix tree. That pgoff_t is usually bitwise-identical to the swp_entry_t. That worked OK when the pagecache was using a hash. But with a radix tree, it produces catastrophically bad results. x86 (and many other architectures) place the `type' field into the low-order bits of the swp_entry_t. So *all* swapcache pages are basically identical in the eight low-order bits. This produces a very sparse radix tree for swapcache. I'm observing packing densities of 1% to 2%: so the typical 128-slot radix tree node has only one or two pages in it. The end result is that the kernel needs to allocate approximately one new radix-tree node for each page which is added to the swapcache. So no wonder we're having radix-tree node exhaustion during swapout! (It's actually quite encouraging that the kernel works as well as it does). The patch changes the encoding of the swp_entry_t so that its most-significant bits contain the `type' field and the least-significant bits contain the `offset' field, right-aligned. That is: the encoding in swp_entry_t is now arch-independent. The new file has conversion functions which convert the swp_entry_t to and from its machine pte representation. Packing density in the swapper_space mapping goes up to around 90% (observed) and the kernel is tons happier under swap load. An alternative approach would be to create new conversion functions which convert an arch-specific swp_entry_t to and from a pgoff_t. I tried that. It worked, but I liked it less. [PATCH] speed up writes Speeds up generic_file_write() by not calling mark_inode_dirty() when the mtime and ctime didn't change. There may be concerns over the fact that this restricts mtime and ctime updates to one-second resolution. But the interface doesn't support that anyway - all the filesystem knows is that its dirty_inode() superop was called. It doesn't know why. So filesystems which support high-resolution timestamps already need to make their own arrangements. We need an update_mtime i_op to support those properly. time to write a one megabyte file one-byte-at-a-time: Before: ext3: 24.8 seconds ext2: 4.9 seconds reiserfs: 17.0 seconds After: ext3: 22.5 seconds ext2: 4.8 seconds reiserfs: 11.6 seconds Not much improvement because we're also calling expensive mark_inode_dirty() functions when i_size is expanded. So compare the overwrite case: time dd if=/dev/zero of=foo bs=1 count=1M conv=notrunc ext3 before: 20.0 seconds ext3 after: 9.7 seconds [PATCH] buffer_boundary() for ext3 Implement buffer_boundary() for ext3. buffer_boundary() is an I/O scheduling hint which the filesystem's get_block() function passes up to the BIO assembly code. It is described in fs/mpage.c The time to read 1,000 52 kbyte files goes from 8.6 seconds down to 2.9 seconds. 52 kbytes is the worst-case size. [PATCH] remove inode.i_wait Remove i_wait from struct inode and hash it instead. This is a pure space-saving exercise - 12 bytes from struct inode on x86. NFS was using i_wait for its own purposes. Add a wait_queue_head_t to the fs-private inode for that. This change has been acked by Trond. [PATCH] rename block_symlink() to page_symlink() block_symlink() is not a "block" function at all. It is a pure pagecache/address_space function. Seeing driverfs calling it was the last straw. The patch renames it to `page_symlink()' and moves it into fs/namei.c [PATCH] direct-to-BIO writeback for writeback-mode ext3 Turn on direct-to-BIO writeback for ext3 in data=writeback mode. [PATCH] give swapper_space a set_page_dirty a_op Give swapper_space a ->set_page_dirty() address_space_operation. So swapcache pages do not need special-casing in set_page_dirty_buffers(). [PATCH] swapcache bugfixes Fixes a few lock ranking bugs (and deadlocks) related to swap_list_lock(), swap_device_lock(), mapping->page_lock and mapping->private_lock. - Cannot call block_flushpage->try_to_free_buffers() inside mapping->page_lock. Because __set_page_dirty_buffers() takes ->page_lock inside ->private-lock. - Cannot call swap_free->swap_list_lock/swap_device_lock inside mapping->page_lock because exclusive_swap_page() takes ->page_lock inside swap_info_get(). The patch also removes all the block_flushpage() calls from the swap code in favour of a direct call to try_to_free_buffers(). The theory is that the page is locked, there is no I/O underway, nobody else has access to the buffers so they MUST be freeable. A bunch of BUG() checks have been added, and unless someone manages to trigger one, the "block_flushpage() inside spinlock" problem is fixed. [PATCH] fix race between writeback and unlink Fixes a race between unlink and writeback: on the sys_sync() and pdflush paths the caller does not have a reference against the inode. So run __iget prior to dropping inode_lock. Oleg Drokin reported this and seems to believe that it fixes the crashes he was observing. But I was never able to reproduce them.. [PATCH] put in-memory filesystem dirty pages on the correct list Replaces SetPageDirty() with set_page_dirty() in several places related to in-memory filesystems. SetPageDirty() is basically always the wrong thing to do. Pages should be moved to the ->dirty_pages list when dirtied so that writeback can see them. Without this change, dirty pages against in-memory filesystems would churn around on the inactive list all the time, rather than getting pushed away onto the active list. A minor efficiency thing. [PATCH] tmpfs bugfixes A patch from Hugh Dickins which fixes a couple of error-path leaks related to tmpfs (I think). Also fixes a yield()-inside-spinlock bug. It also includes code to clear the final page outside i_size on truncate. tmpfs should be returning zeroes when a truncated file is later expanded and it currently is not. Hugh is taking care of the 2.4 fix for this. [PATCH] rename flushpage to invalidatepage Fixes a pet peeve: the identifier "flushpage" implies "flush the page to disk". Which is very much not what the flushpage functions actually do. The patch renames block_flushpage and the flushpage address_space_operation to "invalidatepage". It also fixes a buglet in invalidate_this_page2(), which was calling block_flushpage() directly - it needs to call do_flushpage() (now do_invalidatepage()) so that the filesystem's ->flushpage (now ->invalidatepage) a_op gets a chance to relinquish any interest which it has in the page's buffers. [PATCH] dirsync support for minixfs, sysvfs and ufs Makes minixfs, sysvfs and ufs understand `mount -o dirsync'. [PATCH] 2.5.19 blk.h and more about the ugly kids. - Remove DEVICE_INTR and associated code from floppy driver. - Savlage s390 xpram code from kernel version dependant compilation disease. - Eliminate SET_INTR code from the places where it was used. - Eliminate bogous support for multiple sbpcd controllers. The driver didn't even compile right now before we could think about further supporting it at all we have to get rid of this hack first. Don't call invalidate_buffers in the release method there. Why should it be necessary? - Resurrect sonycd535 compilation. - Let CURRENT request macro use the same primitive at the remaining QUEUE macro in blk.h, which is still not quite right, but first things first :-). idescsi initialization is done with module_init(), not explicitly. Allow to be used even without NULL defined yet. [PATCH] AUDIT 2.5.19: Continuing copy_to/from_user & clear_user Proper usage of isupper/tolower for build scripts. Kernel version 2.5.20