- 28 Oct, 2004 22 commits
-
-
Bjorn Helgaas authored
This preserves the ioreadX() == inX() semantics (same as i386), since it's not clear that the speedup of omitting the mf.a for port operations is worth the risk of breaking drivers. Signed-off-by:
Bjorn Helgaas <bjorn.helgaas@hp.com> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
James Cleverdon authored
Forthcoming IBM boxes will be using Nocona and/or Opteron chips in clustered mode to get beyond 8 CPUs. In fact, there are plans to try for 128 CPUs when the Tulsa chip comes out. Thus, there are a fair number of signed vs. unsigned changes in the patch. Thanks to the HPET timer and some HW changes, I've been able to remove the MPS/ACPI string comparisons from the detection code. Instead, it scans bios_cpu_apicid and uses simple heuristics to select the correct IRQ delivery mode. No need for a config option. Hurrah! Likewise, I've been able to avoid the preprocessor tricks that the i386 sub-arch needed to build with one or more sub-arches. Reluctantly-Acked-by:
Andi Kleen <ak@muc.de> Signed-off-by:
-
Venkatesh Pallipadi authored
hpet hardware seems to need a little prodding during resume for it to start sending the timer interupts again. Attached patch does it for both i386 and x86_64. Makefile change below: Right now suspend-resume ordering of system devices depends on their order of linking. It is ugly. But, thats the way it works currently. And we want timer device to resume before PIC. Signed-off-by:
"Venkatesh Pallipadi" <venkatesh.pallipadi@intel.com> Signed-off-by:
-
Paul Mundt authored
This adds the new syscalls (waitid/add_key/request_key/keyctl). Signed-off-by:
Paul Mundt <paul.mundt@nokia.com> Signed-off-by:
-
Paul Mundt authored
A few random build failures/warnings popped up in 2.6.9, this fixes them. Signed-off-by:
-
Paul Mundt authored
A while back get_signal_to_deliver() changed to take a struct k_sigaction *. This updates do_signal() and handle_signal() to follow the generic API change. Signed-off-by:
-
David Gibson authored
Rework the ppc64 hugepage code. Instead of using specially marked pmd entries in the normal pagetables to represent hugepages, use normal pte_t entries, in a special set of pagetables used for hugepages only. Using pte_t instead of a special hugepte_t makes the code more similar to that for other architecturess, allowing more possibilities for consolidating the hugepage code. Using independent pagetables for the hugepages is also a prerequisite for moving the hugepages into their own region well outside the normal user address space. The restrictions imposed by the powerpc mmu's segment design mean we probably want to do that in the fairly near future. Signed-off-by:
David Gibson <david@gibson.dropbear.id.au> Signed-off-by:
-
Paul Mackerras authored
Unfortunately the patch Ben sent last week to fix a bug in the saving and restoring of floating-point and altivec context across signal handlers introduced another bug, which tends to corrupt the FP and altivec contexts of other tasks. This patch fixes the problem for both ppc32 and ppc64. Signed-off-by:
Paul Mackerras <paulus@samba.org> Signed-off-by:
-
Matt Porter authored
Always disable L2 cache on PPC440GX. All revs/speeds of silicon have parity error problems despite errata claims to the contrary. Signed-off-by:
Matt Porter <mporter@kernel.crashing.org> Signed-off-by:
-
Paul Mackerras authored
This patch is from Nathan Lynch <nathanl@austin.ibm.com>. The NUMA properties of all "possible" cpus are not necessarily available at boot time on ppc64 LPAR. Only the properties for present cpus are known. This patch modifies the ppc64 numa code to map a cpu to its node right before it is brought up -- this means that secondary cpus are now mapped to their nodes during smp_init(). Cpus are removed from their nodes after they have gone offline. Also some minor cleanups: - Stash the "minimum common depth" in a global at boot time, so we don't have to rediscover it every time something changes. - Remove unnecessary variable from of_get_associativity() which is accessed while possibly uninitialized. - Remove the cpu portion from dump_numa_topology() since it will show only the boot cpu now. We could display this information from smp_cpus_done() if necessary. Signed-off-by:
Nathan Lynch <nathanl@austin.ibm.com> Signed-off-by:
-
Stephen Rothwell authored
This patch just removes more of the infrastructure in the PPC64 iSeries console driver that is no longer needed since we no longer need to do copies from user mode in the tty drivers. Signed-off-by:
Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by:
-
James Morris authored
This patch by Kaigai Kohei fixes a bug in the SELinux sidtab code, where we do a spin_unlock_irq() while nested under another irq lock, which enables interrupts and allows a deadlock to happen: sidtab_set() is called between POLICY_WRLOCK and POLICY_WRUNLOCK in services.c:1092. sidtab_set() uses SIDTAB_LOCK()/SIDTAB_UNLOCK(), but SIDTAB_UNLOCK() enables any interruptions because it's defined as spin_unlock_irq(). If an interruption occurs between SIDTAB_UNLOCK() and POLICY_WRUNLOCK, and interruption context try to hold the POLICY_RDLOCK, then a deadlock happen in the result. The solution is to save & restore flags on the inner lock, per the patch below. Signed-off-by:
James Morris <jmorris@redhat.com> Signed-off-by:
Stephen Smalley <sds@epoch.ncsc.mil> Signed-off-by:
Kaigai Kohei <kaigai@ak.jp.nec.com> Signed-off-by:
-
James Morris authored
This patch fixes and simplifies locking in the SELiunux netif cache. An old problem (which I forgot about) is fixed where a netif lookup can be followed by a preemption, causing a race against sel_netif_put(). Kaigai Kohei discovered a problem where netif lookups were also not protected against races with sel_netif_flush(). The code has now been reworked to fix these problems, eliminate the refcounting and remove atomic operations entirely from the read path (generally making better use of RCU). The avc entry ref has been removed as part of this simplification in anticipation of an RCU scalability patch which removes them in general. Signed-off-by:
-
Jeff Garzik authored
Provide a function to get the pageframe number of the nth page at scatterlist.page. We cannot just index off scatterlist.page because the physically-contiguous pages may not be contiguous in mem_map[]. Signed-off-by:
-
Hugh Dickins authored
CRIS does not demand alignment, so PageAnon's PAGE_MAPPING_ANON bit got mixed up with the low bit of the struct address_space *mapping pointer. Patch based on that from Mikael Starvik, but moved the alignment to the declaration of struct address_space itself, and align to sizeof(long) so it's well-aligned on all architectures. Signed-off-by:
Hugh Dickins <hugh@veritas.com> Signed-off-by:
-
Hugh Dickins authored
High latency observed when deleting a large video file from tmpfs. shmem_truncate held info->lock (easily dropped) and nested atomic kmaps (more awkward to handle, since it's structured to keep them) while scanning the many pages of its swap vector. Now be prepared to cond_resched every 64 ops (but scan an empty page in one go). shmem_free_pages to free a linked list of empty swap vector pages at the end (cond_resched every 64): could still free them one by one in the main loop, but this foreshadows scalability changes - which will want to use RCU on them, only freeing the pages after a grace period. Signed-off-by:
-
Hugh Dickins authored
CommitLimit was a good addition to /proc/meminfo, but we don't usually show both what's used and what's free: don't waste lines of screenspace, omit CommitAvail, let the user do the arithmetic as with all the others. And in updating that Documentation, removed the long-gone ReverseMaps. Signed-off-by:
-
Hugh Dickins authored
The sixth "data" field of /proc/$pid/statm was sometimes negative: text is a subset of shared_vm, and so was subtracted twice from total_vm. Signed-off-by:
William Irwin <wli@holomorphy.com> Signed-off-by:
-
Hugh Dickins authored
The third "shared" field of /proc/$pid/statm in 2.4 was a count of pages in the mm whose page_count is more than 1 (oddly, including pages shared just with swapcache). That's too costly to calculate each time, so 2.6 changed it to the total file-backed extent. But Andrea knows apps and users surprised when (rss - shared) goes negative: we need to provide an rss-like statistic, close to the 2.4 interpretation. Something that's quick and easy to maintain accurately is mm->anon_rss, the count of anonymous pages in the mm. Then shared = rss - anon_rss gives a pretty good and meaningful approximation to 2.4's intention: wli confirms that this will be useful to Oracle too. Where to show it? I think it's best to treat this as a bugfix and show it in the third field of /proc/$pid/statm, after resident, as before - there's no evidence that the total file-backed extent was found useful. Albert would like other fields to revert to page counts, but that's a lot harder: if mprotect can change the category of a page, then it can't be accounted as simply as this. Only go that route if real need shown. Signed-off-by:
-
Hugh Dickins authored
The procfs shared_vm accounting in do_mmap_pgoff didn't balance with munmap in the case of shared anonymous: because file comes in NULL, whereas vm_file gets set at the end by shmem_zero_setup. Update file; and update vm_flags (a driver is likely to add VM_IO or VM_RESERVED, modifying reserved_vm); and update pgoff (doesn't affect procfs accounting, but could affect vma_merge - though at present all drivers which modify vm_pgoff set a VM_SPECIAL which prevents merging). And do that __vm_stat_account before advancing to make_pages_present. Signed-off-by:
-
Hugh Dickins authored
The NUMA policy for shared memory or tmpfs page allocation was protected by a semaphore. It helps to improve scalability if we change that to a spinlock (and there's only one place that needs to drop and reacquire). Oh, and don't even bother to get the spinlock while the tree is empty. Acked-by:
Andi Kleen <ak@suse.de> Signed-off-by:
-
David Howells authored
During system boot many probes, etc. will generate upcalls to userspace via call_usermodehelper(). This has the effect of calling execve() before the key subsystem has been initialized, and thus Oopsing on a NULL key_jar during key_alloc(). Move key_init to security_initcall so that it's called along with other security initialization routines which have similar requirements. Signed-off-by:
Chris Wright <chrisw@osdl.org> Signed-off-by:
David Howells <dhowells@redhat.com> Signed-off-by:
-
- 27 Oct, 2004 18 commits
-
-
bk://kernel.bkbits.net/davem/net-2.6Linus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
Herbert Xu authored
Signed-off-by:
Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Herbert Xu authored
Since xfrm_user can be built as a module it's best not to panic when something goes wrong in xfrm_user_init. Call me a wimp for not fixing netlink_kernel_init's return value :) Signed-off-by:
-
Herbert Xu authored
This patch checks the return value of netlink_insert() in netlink_kernel_create(). It could fail if someone loads the same module twice for instance. Signed-off-by:
-
Greg Ungerer authored
This patch fixes 3 problems with the nommu.c support code: 1. export mem_map and vmtruncate (to make them the same as MMU version) 2. update the total_vm usage in mmap routines Without this the OOM killer has no "badness" points to rate processes on... This patch was originally submitted by Giovanni Casoli <giovanni.casoli@atengineering.it> for a 2.4.26 kernel. I applied it to the 2.6.9 code. 3. provide stub for arch_get_unmapped_area Signed-off-by:
Greg Ungerer <gerg@uclinux.org> Signed-off-by:
-
Greg Ungerer authored
Add code to support the apply_relocate_add() function, needed to support all relocation types of the m68k and Coldfire families. This code is straight out of the m68k support for this. Patch was originally from Christian Magnusson <christian.magnusson@runaware.com> Signed-off-by:
Greg Ungerer <gerg@snapgear.com> Signed-off-by:
-
Greg Ungerer authored
Remove remaining unused shglcores.h include. Signed-off-by:
-
Greg Ungerer authored
Remove unused shglcore.h include. Signed-off-by:
-
Greg Ungerer authored
Clean up the definition of HZ on m68knommu systems. There is really only a couple of cases to handle, most boards use the traditional 100Hz still. Only special case out those that don't. The current code is just spaghetti. Signed-off-by:
-
Greg Ungerer authored
Remove unused SHGLCORE support. Signed-off-by:
-
Greg Ungerer authored
Remove uneeded includes. asm/delay.h is not needed here. Signed-off-by:
-
Greg Ungerer authored
Remove uneeded includes. asm/delay.h is not needed here. Signed-off-by:
-
Greg Ungerer authored
Remove uneeded includes. asm/delay.h is not needed here. Signed-off-by:
-
Greg Ungerer authored
Remove unused shglcore.h include from m68knommu/mm/memory.c That file will be removed in an upcoming patch. Signed-off-by:
-
Greg Ungerer authored
Remove unused shglcore.h include from m68knommu/mm/init.c That file will be removed in an upcoming patch. Signed-off-by:
-
Linus Torvalds authored
Kbuild can't fix up "-I $(src)/.." to the right directory due to the particular magic transformations it does on -I arguments. Remove the space, use "-I$(src)/.." and it works. Hopefully Sam will fix this build issue. I think somebody should dress up as scripts/Makefile.lib for Halloween. Now _that_ would be scary.
-
Linus Torvalds authored
Noted by Al Viro, superhacker and endlessly patient person. Sign up today! You too can become a superhacker by setting up a cross-compile environment for all architectures and building "allmodconfig" trees all day long to find places where others have screwed up.
-
Randolph Chung authored
If an arch Makefile overrides CROSS_COMPILE (e.g. parisc, mips, ...) then the cc-option call in the main Makefile uses the wrong compiler to check for options. Signed-off-by:
Randolph Chung <tausq@debian.org> Signed-off-by:
-
