- 15 Jan, 2005 40 commits
-
-
Anton Blanchard authored
With David Woodhouse <dwmw2@infradead.org> We don't set MM_VM_SIZE() on ppc64, so it defaults to TASK_SIZE. Which means a 32-bit process ending up in exit_mmap() to kill a 64-bit mm may call tlb_finish_mmu() with an incorrect 'end' argument. Signed-off-by:
Anton Blanchard <anton@samba.org> Signed-off-by:
David Woodhouse <dwmw2@infradead.org> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Olof Johansson authored
On some systems, the first PCI bus has a ISA I/O hole at the first 16MB. We can't use this space for DMA addresses on the bus. On Python-based machines, we'll skip the first 256MB on buses that have the hole, just as we do on later systems. This means that the first bus will have 768MB of DMA space shared between the devices on it. Signed-off-by:
Olof Johansson <olof@austin.ibm.com> Acked-by:
Paul Mackerras <paulus@samba.org> Signed-off-by:
-
Michael Ellerman authored
When the iseries_veth driver module is unloaded there is the potential for an oops and also some memory leakage. Because the HvLpEvent_unregisterHandler() function did no synchronisation, it was possible for the handler that was being unregistered to be running on another CPU *after* HvLpEvent_unregisterHandler() had returned. This could cause the iseries_veth driver to leave work in the events work queue after the module had been unloaded. When that work was eventually executed we got an oops. In addition some of the data structures in the iseries_veth driver were not being correctly freed when the module was unloaded. This is the second patch, we make iseries_veth call flush_scheduled_work() after we are sure the handler is no longer running, and also fix the memory leaks. Signed-off-by:
Michael Ellerman <michael@ellerman.id.au> Signed-off-by:
-
Michael Ellerman authored
When the iseries_veth driver module is unloaded there is the potential for an oops and also some memory leakage. Because the HvLpEvent_unregisterHandler() function did no synchronisation, it was possible for the handler that was being unregistered to be running on another CPU *after* HvLpEvent_unregisterHandler() had returned. This could cause the iseries_veth driver to leave work in the events work queue after the module had been unloaded. When that work was eventually executed we got an oops. In addition some of the data structures in the iseries_veth driver were not being correctly freed when the module was unloaded. This is the first patch, which makes HvLpEvent_unregisterHandler() work. Signed-off-by:
-
Matt Porter authored
Fix PPC44x build broken from a latent bug. Signed-off-by:
Matt Porter <mporter@kernel.crashing.org> Signed-off-by:
-
Tom Rini authored
The following adds support for Freescale's PQ2FADS board to the kernel. (The 'real' changes required for the board are in the ethernet driver, which needs a massive sync-up with mainline as I've been holding out for a better, rewritten one from the Freescale folks, who're waiting for the PHY lib patch sent to the net-dev list to get ack'd) Signed-off-by:
Tom Rini <trini@kernel.crashing.org> Signed-off-by:
-
Tom Rini authored
When I reworked the m82xx init functions, I inadvertantly broke the callout we had from ppc_md.setup_arch() that boards can use to poke & prod things, once mappings are set. The following adds in a callback and updates the one m82xx board that needs it. Signed-off-by:
-
Miquel van Smoorenburg authored
I just discovered there's a thinko in the mark-page-accessed change in do_generic_mapping_read() in 2.6.11-rc1. ra.prev_page is compared to index to see if we read from this page before - except that prev_page is actually set to the recent page or even a page in front of the current page. So we should store ra.prev_page in a seperate variable at the start of do_generic_mapping_read(). Signed-off-by:
-
Roland Dreier authored
Change routine name from handle_outgoing_smp to handle_outgoing_dr_smp. Signed-off-by:
Hal Rosenstock <halr@voltaire.com> Signed-off-by:
Roland Dreier <roland@topspin.com> Signed-off-by:
-
Roland Dreier authored
Move the gather list and work request used for posting sends from the stack in ipoib_send() to the private structure. This reduces the stack usage for the data path function ipoib_send() and may speed things up slightly because we don't need to initialize constant members of the structures. Signed-off-by:
-
Roland Dreier authored
Update copyright line (files were modified in 2005). Signed-off-by:
-
Roland Dreier authored
Add a new function to find a port on a device given a GID by searching the cached GID tables. Document all cache functions in ib_cache.h. Rename existing functions to better match format of verb routines. Signed-off by: Sean Hefty <sean.hefty@intel.com> Signed-off-by:
-
Roland Dreier authored
Add qp_type to struct ib_qp. Signed-off by: Sean Hefty <sean.hefty@intel.com> Signed-off-by:
-
Roland Dreier authored
Add parameters to process_mad device method to support full Mellanox firmware capabilities (pass sufficient information for baseboard management trap generation, etc). Signed-off-by:
Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by:
-
Roland Dreier authored
Handle outgoing DR 0 hop SMPs properly when provider returns just SUCCESS to process_mad. Signed-off-by:
-
Roland Dreier authored
Clean up the computation of the HCA context memory map. This serves two purposes: - make it easier to change the HCA "profile" (eg add more QPs) - make it easier to implement mem-free Arbel support Signed-off-by:
-
Roland Dreier authored
Add per-device "node_type" and per-port "phys_state" sysfs attributes for InfiniBand devices. Signed-off-by:
-
Roland Dreier authored
InfiniBand spec rev 1.2 compliance: add local qp number to work completion structure. Signed-off-by:
-
Roland Dreier authored
Integrate Michael Tsirkin's patch to local_completion to set the WC byte_cnt according to the IBA 1.1 spec (include the GRH size regardless of whether it is present or not). Signed-off-by:
-
Roland Dreier authored
Make needlessly global code static. Signed-off-by:
Adrian Bunk <bunk@stusta.de> Signed-off-by:
-
Roland Dreier authored
Remove debug printk accidentally included. Signed-off-by:
Tom Duffy <tduffy@sun.com> Signed-off-by:
-
Roland Dreier authored
Add an rmb() between checking the ownership bit of an event queue entry and reading the contents of the EQE. Without this barrier, the CPU could read stale contents of the EQE before HW writes the EQE but have the read of the ownership bit reordered until after HW finishes writing, which leads to the driver processing an incorrect event. This was actually observed to happen when multiple completion queues are in heavy use on an IBM JS20 PowerPC 970 system. Also explain the existing rmb() in completion queue poll (there for the same reason) and slightly improve debugging output. Signed-off-by:
-
Roland Dreier authored
Clean up the way we allocate and map memory for use as ICM ("InfiniHost Context Memory") when running in Arbel MemFree mode. This slightly improves the code for mapping the firmware area and will make future progress towards full MemFree support much easier. Signed-off-by: -
Roland Dreier authored
Implement setting of RDMA/atomic enable bits, initiator resources and responder resources for modify QP in low-level Mellanox HCA driver (should complete RDMA/atomic implementation). Signed-off-by:
-
Roland Dreier authored
Trivial formatting fix for empty for loops. Signed-off-by:
-
Roland Dreier authored
Calculate static rate for IPoIB address handles based on local width/speed and path rate. Signed-off-by:
-
James Bottomley authored
Matthew Wilcox just converted parisc over to doing the generic irq code and we ran across the symbol probe_irq_mask being undefined (and thus preventing yenta_socket from loading). It looks like the EXPORT_SYMBOL() was accidentally missed from kernel/irq/autoprobe.c and no-one noticed on x86 because it's still in i386_ksyms.c This patch corrects the problem so that the generic irq code now works completely on parisc. Signed-off-by:
James Bottomley <James.Bottomley@SteelEye.com> Signed-off-by:
-
Stephen D. Smalley authored
This patch fixes a different bug in the code for SELinux policy loading. It ensures that the loaded policy version number is not updated until the new policy is successfully committed. It also fixes the type on the loaded policy version. Signed-off-by:
Stephen Smalley <sds@epoch.ncsc.mil> Signed-off-by:
-
Stephen D. Smalley authored
This patch fixes several bugs in the error handling code for SELinux policy loading that were introduced by my earlier patch to eliminate unaligned accesses by that code. Signed-off-by:
-
Ulrich Weigand authored
account_steal_time called for idle doesn't work correctly: 1) steal time while idle needs to be added to the system time of idle to get correct uptime numbers 3) if there is an i/o request outstanding the steal time should be added to iowait, even if the hypervisor scheduled another virtual cpu since we are still waiting for i/o. 2) steal time while idle without an i/o request outstanding has to be added to cpustat->idle and not to cpustat->system. Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com> Acked-by:
Ingo Molnar <mingo@elte.hu> Signed-off-by:
-
bk://linux-sam.bkbits.net/kconfigLinus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
Linus Torvalds authored
Use a "pipe_buf_operations" structure to describe the ops that can be done on a pipe buffer, so that pipe buffers from different sources can have their own rules. Right now the rules are just about how you map the buffers into kernel virtual memory space, and how to release them (and whether you can append new data to the end of an existing buffer).
-
bk://linux-dj.bkbits.net/cpufreqLinus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
Matthew Wilcox authored
Compat syscalls need to start compat_sys_ otherwise PA-RISC's compat syscall wrappers don't work. Not that the individual involved bothered to patch PA-RISC ... Signed-off-by: -
Ingo Molnar authored
Paul Mackerras points out that doing the _raw_spin_trylock each time through the loop will generate tons of unnecessary bus traffic. Instead, after we fail to get the lock we should poll it with simple loads until we see that it is clear and then retry the atomic op. Assuming a reasonable cache design, the loads won't generate any bus traffic until another cpu writes to the cacheline containing the lock. Agreed. Signed-off-by:
-
-
Bartlomiej Zolnierkiewicz authored
Device drivers are supposed to provide their own ->attach and ->cleanup. Signed-off-by:Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
-
Bartlomiej Zolnierkiewicz authored
Use now available ide_dma_intr() in etrax_dma_intr(). Signed-off-by: -
Bartlomiej Zolnierkiewicz authored
Make ide_dma_intr() always available if CONFIG_BLK_DEV_IDEDMA=y and use it instead of icside_dmaintr(). Acked by Russell. Signed-off-by: -
Bartlomiej Zolnierkiewicz authored
This cruft pre-dates benh's PowerManagement code. Moreover ->suspend_reset is always equal to zero. Signed-off-by:
-
