The main news of 2.4.5pre5aa1 are the alpha updates in particular the iommu fixes that are supposed to stablize the big alpha boxes in production. The most important bit may be still missing from this diff because I want to force people to test a possible alternate "real" fix for the iommu instability. Maybe the only bug that had to be fixed is the one addressed by 00_alpha-iommu-tlb-corruption-1 that I spotted and fixed today. So in short if 2.4.5pre5aa1 still crashes with tons of ram on the tsunami chipsets, then you can apply this incremental patch on top of 2.4.5pre5aa1 and it should stabilize again: ftp://ftp.kernel.org/pub/linux/kernel/people/andrea/patches/v2.4/2.4.5pre5aa1/tsunami-iommu-align-1 I just want to make 100% sure that the above patch introducing the artifical alignment to avoid invalid (but not corrupted) pte to be cached in the tlb is necessary and that the real bug wasn't the corrupted ptes generated during pci_map_* and fixed by 00_alpha-iommu-tlb-corruption-1. --------------------------------------------------------------------------- Detailed description on the diff between 2.4.5pre2aa1 and 2.4.5pre5aa1 follows. --------------------------------------------------------------------------- Only in 2.4.5pre5aa1: 00_alpha-cia-1 Fix for CIA by Ivan plus a few additional fixes. (recommended) Only in 2.4.5pre5aa1: 00_alpha-cypress-quirk-1 Never use pci32 addresses aboe -1M to avoid confusing cypress chip. (recommended) Only in 2.4.5pre5aa1: 00_alpha-dp264-compile-1 Allows non generic dp264 compiles to link. (nice to have) Only in 2.4.5pre5aa1: 00_alpha-iommu-locking-1 Fixes races in the pci_unmap_* paths of alpha. (recommended) Only in 2.4.5pre5aa1: 00_alpha-iommu-printk-1 Convert to warnings printk notifying about pci_map_* faliures (should be KERN_CRIT but think positive and assume we'll fix the bugs eventually ;). (nice to have) Only in 2.4.5pre5aa1: 00_alpha-iommu-tlb-corruption-1 Avoid corrupting the tlb and possibly crash the iommu with pagetables bits that must be zero. (recommended) Only in 2.4.5pre2aa1: 00_alpha-numa-10 Only in 2.4.5pre5aa1: 00_alpha-numa-11 Minor changes in the alpha-numa support, compilation issues. (update) Only in 2.4.5pre5aa1: 00_alpha-sg-direct-1 Fix for the patch from Richard that uses the direct window during SG mappings if we run out of pci dynamic mappings (by the time we run out of dynamic mappings we're going to break very soon anyways, but what this patch does is certainly a good idea). (recommended) Only in 2.4.5pre5aa1: 00_alpha-spinlock-debug-tmo-1 Increase the spinlock debugging timeout trigger for the faster cpus. (spinlock debugging is disabled anyways) (nice to have) Only in 2.4.5pre5aa1: 00_alpha-tsunami-1G-dynamic-pci-1 Increase the dynamic window for tsunami from 128M to 1G (make sure to apply also the 00_alpha-cypress-quirk-1 if you apply this by hand). This tries to better hide the bugs in the drivers that doesn't check for pci_map_* faliures. (recommended) Only in 2.4.5pre5aa1: 00_alpha-tsunami-no-wmb-1 Drop a not necessary wmb(). (nice to have) Only in 2.4.5pre5aa1: 00_alpha-tsunami-pci64-enable-1 Enable monster window on the tsunami chip plus define TSUNAMI_DAC_OFFSET. This is a backdoor to allow device drivers that are aware they are running on a tsunami chipset to use DAC on pci64. What the driver needs to do in order to convert a physical address to a pci64 bus address is to `logical or' the physical address with the TSUNAMI_DAC_OFFSET (that is 1<<40), and then program the chip to start the dma on the so calculated 64bit bus address. This is a very primitve backdoor, implementing a virt_to_bus64 isn't that far and a virt_to_bus64 would be enough for alpha, plus in any case we'll also need a common code abstraction to allow the driver to know if it can use DAC or not, which is missing as well at the moment. It is recommended to use this backdoor only to decrease the pressure on the iommu mappings, so that we decrease the probability of triggering all the device driver bugs. (nice to have for most, recommended for anybody running out of entries with a pci64 device on a tsunami chip) Only in 2.4.5pre5aa1: 00_boot-serial-console-1 Allows the serial console to work anytime during boot. It may have side effects but certainly nothing relevant and the current situation was annoying enough. (nice to have) Only in 2.4.5pre5aa1: 00_eepro100-64bit-1 Fixes a 64bit bug that was generating false positives and memory corruption. (recommended) Only in 2.4.5pre2aa1: 00_nfs-corruption-3 Dropped the flush at unmap thing, replaced with the flush at release from mainline. NOTE: the 00_nfs-corruption-3 was not buggy for production, this is just an alternate better way of fixing the corruption. Only in 2.4.5pre2aa1: 00_rwsem-11 Only in 2.4.5pre5aa1: 00_rwsem-13 Updated to compile. I had to disable ppc asm optimizations as I didn't had time to look into them. Unless somebody sends me a patch I will look into it shortly (once I return in optimization mode, now I am in stability mode ;). (update) Only in 2.4.5pre5aa1: 00_xircom-serial-1 Merged a fix posted on l-k that makes xircom integrated modem to work fine (without this artificial delay during initialization it really doesn't work). (nice to have) Only in 2.4.5pre2aa1: 00_xircom-tulip-cb-arjanv-1.bz2 Only in 2.4.5pre5aa1: 00_xircom-tulip-cb-arjanv-2.bz2 Update with the latest driver at http://people.redhat.com/arjanv --------------------------------------------------------------------------- Detailed description of 2.4.5pre2aa1 follows. --------------------------------------------------------------------------- 00_alpha-illegal-irq-1 Be verbose for MAX_ILLEGAL_IRQS times if an invalid irq number is getting run. (debugging) 00_alpha-ksyms-1 Export a few alpha-arch symbols needed by modules. (recommended to avoid compilation troubles) 00_alpha-large-vmalloc-1 Drop the CONFIG_LARGE_VMALLOC selection from the arch/alpha/config.in, the large-vmalloc feature is racy and it can destabilize the machine, fixing it isn't worthwhile because nobody needs more than 8Gigabytes of ram in vmalloc memory, not even tux on the 256G boxes will ever need that. (recommended) 00_alpha-modrace-1 Fix alpha races between module insmod/rmmod and the page fault fixmap lookup. (recommended) 00_alpha-numa-10 Fully support wildfire machines with all kind of NUMA memory configuration, plus it optimizes the allocation on per node basis to boost the performance on the NUMA boxes. Right now CONFIG_WILDFIRE needs to be selected to take advantage of this feature. (CONFIG_GENERIC + CONFIG_DISCONTIGMEM=y and CONFIG_NUMA=y will work fine as well but it won't take advantage of the new feature. It also fixes many memory management bit in the core linux allocator in the common code, mostly to avoid wasting static memory. (recommended) 00_alpha-sched-yield-1 Fixes SCHED_YIELD on the alpha arch for UP compiles. (recommended) 00_alpha-show_stack-1 Implements the show_stack() call used often by some common code, mostly to allow compilation, things like tux needs it. (nice to have) 00_alpha-tlb-page-sym-1 Drops a not necessary export on the alpha port. (recommended) 00_buffer-2 Reschedule during oom while allocating buffers, still getblk can deadlock with oom but this will hide it pretty well as it won't loop in a tight loop anymore. (recommended) 00_cachelinealigned-in-smp-1 Moves the pagecache_lock and the VM pagemap_lru_lock in two different L1 cachelines to avoid contention, mostly useful on the alpha where the spinlocks uses load locked store conditional loops (and we don't want to loop). (nice to have) 00_copy-user-lat-2 Put the rechedule points into copy-user calls, with lots of cache large read/writes could otherwise _never_ reschedule once until they returns to userspace. (recommended) 00_cpus_allowed-1 Fixes a bug in the cpu affinity in-kernel API, bug was fatal for ksoftirqd. (recommended) 00_double-buffer-pass-1 Avoids looping two times for no good reason into the lru lists of the buffer cache (the double loop was an unreliable hack from the prehistory that survided 'till today). (nice to have) 00_exception-table-1 Avoids a compilation warning when compiling without modules. (very minor thing) 00_highmem-deadlock-3 Fixes an highmem deadlock using a reserved pool for the bounce buffers. (recommended) 00_highmem-debug-1 Allows people with x86 machines with less than 1G of ram to test the highmem code. (debugging) 00_ia32-bootmem-corruption-1 Fixes the x86 boot stage to finish initializing all the reserved memory before starting allocating memory. (recommended) 00_ipv6-null-oops-1 Fixes null pointer oops. (recommended) 00_jens-loop-noop-nobounce-1 Skips the bounces with the null transfer function. (nice to have) 00_ksoftirqd-4 Avoids 1/HZ latency for the softirq if the softirq is marked again pending when do_softirq() finished and the machine is otherwise idle, it also fixes the case of a softirq re-marking itself runnable by delegating to the scheduler the balance of the softirq load like if it would be an normal task. (nice to have) 00_kupdate-large-interval-1 Allows to set large interval for the kupdate runs, this is useful on the laptops, instead of sigstopping ksoftirqd it's nicer to set a large interval for example of the order of one hour (do that at your own risk of course, doing that is not recommended unless you know what you're doing). (nice to have) 00_lvm-0.9.1_beta7-4 Updates to the lvmbeta7 with fixes for the lv hardsectsize estimantion based on the max hardsectsize of the underlying pv, plus it has some other tons of fixes and it is a must have for the 64bit archs as the IOP silenty changed for those platforms. (recommended) 00_max_readahead-1 Increases the max_readahead to allow the blkdev to read with 512k scsi commands when possible. (nice to have) 00_msync-fb0-1 Fixes oopses while running msync on a region of virtual memory that maps to reserved memory. (recommended) 00_nfs-corruption-3 Production fix for the nfs map_shared fs data corruption. Other design solutions are discussed and the long term fix might be different but all the other approches are more invasive and risky, while this one is obviously right and the most approriate for production. (recommended) 00_numa-sched-6 Implements a basic numa scheduler with a per node runqueue that boosts the perforormance on numa boxes. It also enables and somewhere fixes the last_idle heuristic in the smp scheduler. The numa part doesn't impose any runtime overhead when CONFIG_NUMA_SCHED is not set. (nice to have) 00_o_direct-6 Implements O_DIRECT zerocopy direct I/O in a filesystem indipendent fascion, currently it's only available on ext2, but it will be easy to let the other of fs to take advantage of it too. (nice to have) 00_pagetable-fast-2 Enables the usage of the per-cpu quicklists for the pte/pgd to optimize the cache affinity and footprint. (nice to have) 00_peekurgdata-1 Allows MSG_PEEK to work on the urgent (aka out of band) receive queue as well, needed for tux. (nice to have) 00_rwsem-11 Alternate implementation for the rwsem, the x86 asm version is simpler in the slow path and it provides a faster up_write fast path, the C-spinlock version is much faster too. (nice to have) 00_sched-yield-1 Fixes a bug in sched_yield, where SCHED_YIELD needs to be cleared after a schedule() if no other task was runnable, otherwise the next schedule() would inherit the SCHED_YIELD behaviour even if a SCHED_YIELD wasn't requested anymore. (nice to have) 00_show_regs-1 Makes SYSRQ+P more verbose. (debugging) 00_slab-lists-1 Rewrites the slab cache handling for partial and completly free slab objects, always provides LIFO behaviour from all the lists to reduce the cache misses, while previously the completly free slab objects were reallocated with a less optimal FIFO policy. It also cleanups the code. (nice to have) 00_softirq-SMP-fixes-3 Fixes an SMP race in the softirq code on archs like the alpha where the atomic_t and bitop operations aren't memory barriers as well. (recommended) 00_sync-page-1 Avoids suprious unplug of the tq_disk task queue while waiting I/O completion. (nice to have) 00_timer_t-2 Defines the timer_t type in one single place. (nice to have) 00_waitqueue-2 Setups the not yet visible waitqueue flags outside the critical section. (nice to have) 00_x86-systable-1 Fills the end of the syscall table automatically, it is off by one without this patch. (nice to have) 00_xircom-tulip-cb-arjanv-1 No need of ifconfig eth0 promisc on the xircom card with this driver. (nice to have) 10_no-virtual-1 Avoids wasting tons of memory if highmem is not selected (like in all the 64bit ports). (nice to have) 10_parent-timeslice-7 Fixes a scheduler unfairness generated by the parent-timeslice logic. (recommended) 10_read_ahead-2 Setups readahead for lvm [global, hacking around read/write callbacks is broken] and drops sensless limits. (recommended) 20_share-timeslice-2 Reschedule the child first but share the timeslice with the parent (I never triggered a single userspace bug with this correct patch) (nice to have) The below patches aren't included by default but they can be applied incrementally if tux is needed. TUX patches are been developed by Ingo Molnar. 30_atomic-alloc-1 Defines a PF_ATOMICALLOC that won't sleep watiting memory to become available. (needed by tux) 30_atomic-lookup-1 Implements an O_ATOMICLOOKUP flag to avoid entering the filesystem code that could wait for I/O in open, open will return -EWOULDBLOCKIO if the lookup wasn't doable in dcache. (needed by tux for the asynchronous vfs lookups) 30_net-exports-1 Exports some networking function so that tux can be compiled as module, and splits the sock_map_file functionality away from sock_map_fd. (needed by tux) 30_pagecache-atomic-1 Defines a do_generic_file_read_atomic that implements nonblocking reads from cache, if information wasn't in cache the descriptor error flag is set to -EWOULDBLOCKIO. (needed by tux to implement asynchronous read I/O) 30_tux-1 tux core. (nice to have) 30_tux-data-1 Adds a tux_data private pointer to the sock structure. (needed by tux) 30_tux-dprintk-1 Defines tux_Dprintk. (needed by tux) 30_tux-exports-1 Exports to modules the in-kernel syscalls. (needed by tux) 30_tux-kstat-1 Defines additional kstats for tux. 30_tux-process-1 Defines per-process callbacks for tux. 30_tux-syscall-1 Implements the tux-syscall, it's the one executed by the userspace tux program to fire up, shutdown and control the tux kernel threads. (needed by tux) 30_tux-sysctl-1 Defines the symbolic names for the tux sysctl. (needed by tux) 30_tux-vfs-1 Adds a per-dentry tux private data. (needed by tux) 31_tux-logger-1 Aligns correctly the tux logentry for 8 byte cachelines too. --------------------------------------------------------------------------- Have fun! Andrea