8 years agoi915/gtt: fix ordering issues with status setup and DMAR
Dave Airlie [Wed, 12 Jan 2011 01:38:37 +0000]
i915/gtt: fix ordering issues with status setup and DMAR

This code was setting up the status page before setting the DMAR-is-on-bit,
so we were getting DMAR errors on the status page. Reverse the two bits
of init code to the correct result.

Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agoMerge branch 'drm-intel-fixes' of ssh://master.kernel.org/pub/scm/linux/kernel/git...
Dave Airlie [Wed, 12 Jan 2011 00:49:03 +0000]
Merge branch 'drm-intel-fixes' of ssh:///linux/kernel/git/ickle/drm-intel

* 'drm-intel-fixes' of ssh://master.kernel.org/pub/scm/linux/kernel/git/ickle/drm-intel: (37 commits)
  drm/i915/execbuffer: Reorder binding of objects to favour restrictions
  drm/i915: If we hit OOM when allocating GTT pages, clear the aperture
  drm/i915/evict: Ensure we completely cleanup on failure
  drm/i915/execbuffer: Correctly clear the current object list upon EFAULT
  drm/i915/debugfs: Show all objects in the gtt
  drm/i915: Record AGP memory type upon error
  drm/i915: Periodically flush the active lists and requests
  drm/i915/gtt: Unmap the PCI pages after unbinding them from the GTT
  drm/i915: Record the error batchbuffer on each ring
  drm/i915: Include TLB miss overhead for computing WM
  drm/i915: Propagate error from flushing the ring
  drm/i915: detect & report PCH display error interrupts
  drm/i915: cleanup rc6 code
  drm/i915: fix rc6 enabling around suspend/resume
  drm/i915: re-enable rc6 support for Ironlake+
  drm/i915: Make the ring IMR handling private
  drm/i915/ringbuffer: Simplify the ring irq refcounting
  drm/i915/debugfs: Show the per-ring IMR
  drm/i915: Mask USER interrupts on gen6 (until required)
  drm/i915: Handle ringbuffer stalls when flushing
  ...

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Wed, 12 Jan 2011 00:32:41 +0000]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (67 commits)
  cxgb4vf: recover from failure in cxgb4vf_open()
  netfilter: ebtables: make broute table work again
  netfilter: fix race in conntrack between dump_table and destroy
  ah: reload pointers to skb data after calling skb_cow_data()
  ah: update maximum truncated ICV length
  xfrm: check trunc_len in XFRMA_ALG_AUTH_TRUNC
  ehea: Increase the skb array usage
  net/fec: remove config FEC2 as it's used nowhere
  pcnet_cs: add new_id
  tcp: disallow bind() to reuse addr/port
  net/r8169: Update the function of parsing firmware
  net: ppp: use {get,put}_unaligned_be{16,32}
  CAIF: Fix IPv6 support in receive path for GPRS/3G
  arp: allow to invalidate specific ARP entries
  net_sched: factorize qdisc stats handling
  mlx4: Call alloc_etherdev to allocate RX and TX queues
  net: Add alloc_netdev_mqs function
  caif: don't set connection request param size before copying data
  cxgb4vf: fix mailbox data/control coherency domain race
  qlcnic: change module parameter permissions
  ...

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Wed, 12 Jan 2011 00:32:20 +0000]
Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  sparc64: Fix bootup regression due to perf init ordering.

8 years agoMerge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Wed, 12 Jan 2011 00:31:41 +0000]
Merge branch 'next' of git://git./linux/kernel/git/benh/powerpc

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc: (72 commits)
  powerpc/pseries: Fix build of topology stuff without CONFIG_NUMA
  powerpc/pseries: Fix VPHN build errors on non-SMP systems
  powerpc/83xx: add mpc8308_p1m DMA controller device-tree node
  powerpc/83xx: add DMA controller to mpc8308 device-tree node
  powerpc/512x: try to free dma descriptors in case of allocation failure
  powerpc/512x: add MPC8308 dma support
  powerpc/512x: fix the hanged dma transfer issue
  powerpc/512x: scatter/gather dma fix
  powerpc/powermac: Make auto-loading of therm_pm72 possible
  of/address: Use propper endianess in get_flags
  powerpc/pci: Use printf extension %pR for struct resource
  powerpc: Remove unnecessary casts of void ptr
  powerpc: Disable VPHN polling during a suspend operation
  powerpc/pseries: Poll VPA for topology changes and update NUMA maps
  powerpc: iommu: Add device name to iommu error printks
  powerpc: Record vma->phys_addr in ioremap()
  powerpc: Update compat_arch_ptrace
  powerpc: Fix PPC_PTRACE_SETHWDEBUG on PPC_BOOK3S
  powerpc/time: printk time stamp init not correct
  powerpc: Minor cleanups for machdep.h
  ...

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Wed, 12 Jan 2011 00:30:08 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (42 commits)
  IB/qib: Fix refcount leak in lkey/rkey validation
  IB/qib: Improve SERDES tunning on QMH boards
  IB/qib: Unnecessary delayed completions on RC connection
  IB/qib: Issue pre-emptive NAKs on eager buffer overflow
  IB/qib: RDMA lkey/rkey validation is inefficient for large MRs
  IB/qib: Change QPN increment
  IB/qib: Add fix missing from earlier patch
  IB/qib: Change receive queue/QPN selection
  IB/qib: Fix interrupt mitigation
  IB/qib: Avoid duplicate writes to the rcv head register
  IB/qib: Add a few new SERDES tunings
  IB/qib: Reset packet list after freeing
  IB/qib: New SERDES init routine and improvements to SI quality
  IB/qib: Clear WAIT_SEND flags when setting QP to error state
  IB/qib: Fix context allocation with multiple HCAs
  IB/qib: Fix multi-Florida HCA host panic on reboot
  IB/qib: Handle transitions from ACTIVE_DEFERRED to ACTIVE better
  IB/qib: UD send with immediate receive completion has wrong size
  IB/qib: Set port physical state even if other fields are invalid
  IB/qib: Generate completion callback on errors
  ...

8 years agopowerpc/pseries: Fix build of topology stuff without CONFIG_NUMA
Benjamin Herrenschmidt [Tue, 11 Jan 2011 23:56:29 +0000]
powerpc/pseries: Fix build of topology stuff without CONFIG_NUMA

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

8 years agocxgb4vf: recover from failure in cxgb4vf_open()
Casey Leedom [Tue, 11 Jan 2011 23:44:40 +0000]
cxgb4vf: recover from failure in cxgb4vf_open()

If the Link Start fails in cxgb4vf_open(), we need to back out any state
that we've built up ...

Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoMerge branch 'master' of git://1984.lsi.us.es/net-2.6
David S. Miller [Tue, 11 Jan 2011 23:43:03 +0000]
Merge branch 'master' of git://1984.lsi.us.es/net-2.6

8 years agoMerge branch 'nfs-for-2.6.38' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
Linus Torvalds [Tue, 11 Jan 2011 23:11:56 +0000]
Merge branch 'nfs-for-2.6.38' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6

* 'nfs-for-2.6.38' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6: (89 commits)
  NFS fix the setting of exchange id flag
  NFS: Don't use vm_map_ram() in readdir
  NFSv4: Ensure continued open and lockowner name uniqueness
  NFS: Move cl_delegations to the nfs_server struct
  NFS: Introduce nfs_detach_delegations()
  NFS: Move cl_state_owners and related fields to the nfs_server struct
  NFS: Allow walking nfs_client.cl_superblocks list outside client.c
  pnfs: layout roc code
  pnfs: update nfs4_callback_recallany to handle layouts
  pnfs: add CB_LAYOUTRECALL handling
  pnfs: CB_LAYOUTRECALL xdr code
  pnfs: change lo refcounting to atomic_t
  pnfs: check that partial LAYOUTGET return is ignored
  pnfs: add layout to client list before sending rpc
  pnfs: serialize LAYOUTGET(openstateid)
  pnfs: layoutget rpc code cleanup
  pnfs: change how lsegs are removed from layout list
  pnfs: change layout state seqlock to a spinlock
  pnfs: add prefix to struct pnfs_layout_hdr fields
  pnfs: add prefix to struct pnfs_layout_segment fields
  ...

8 years agonetfilter: ebtables: make broute table work again
Florian Westphal [Tue, 11 Jan 2011 22:55:51 +0000]
netfilter: ebtables: make broute table work again

broute table init hook sets up the "br_should_route_hook" pointer,
which then gets called from br_input.

commit a386f99025f13b32502fe5dedf223c20d7283826
(bridge: add proper RCU annotation to should_route_hook)
introduced a typedef, and then changed this to:

br_should_route_hook_t *rhook;
[..]
rhook = rcu_dereference(br_should_route_hook);
if (*rhook(skb))

problem is that "br_should_route_hook" contains the address of the function,
so calling *rhook() results in kernel panic.

Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

8 years agodrm/i915/execbuffer: Reorder binding of objects to favour restrictions
Chris Wilson [Mon, 10 Jan 2011 17:35:37 +0000]
drm/i915/execbuffer: Reorder binding of objects to favour restrictions

As the mappable portion of the aperture is always a small subset at the
start of the GTT, it is allocated preferentially by drm_mm. This is
useful in case we ever need to map an object later. However, if you have
a large object that can consume the entire mappable region of the
GTT this prevents the batchbuffer from fitting and so causing an error.
Instead allocate all those that require a mapping up front in order to
improve the likelihood of finding sufficient space to bind them.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: If we hit OOM when allocating GTT pages, clear the aperture
Chris Wilson [Mon, 10 Jan 2011 17:33:15 +0000]
drm/i915: If we hit OOM when allocating GTT pages, clear the aperture

Rather than evicting an object at random, which is unlikely to alleviate
the memory pressure sufficient to allow us to continue, zap the entire
aperture. That should give the system long enough to recover and reap
some pages from the evicted objects, forestalling the allocation error
for the new object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915/evict: Ensure we completely cleanup on failure
Chris Wilson [Mon, 10 Jan 2011 14:21:05 +0000]
drm/i915/evict: Ensure we completely cleanup on failure

... and not leave the objects in a inconsistent state.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

8 years agodrm/i915/execbuffer: Correctly clear the current object list upon EFAULT
Chris Wilson [Mon, 10 Jan 2011 12:09:12 +0000]
drm/i915/execbuffer: Correctly clear the current object list upon EFAULT

Before releasing the lock in order to copy the relocation list from user
pages, we need to drop all the object references as another thread may
usurp and execute another batchbuffer before we reacquire the lock.
However, the code was buggy and failed to clear the list...

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

8 years agonetfilter: fix race in conntrack between dump_table and destroy
Stephen Hemminger [Tue, 11 Jan 2011 22:54:42 +0000]
netfilter: fix race in conntrack between dump_table and destroy

The netlink interface to dump the connection tracking table has a race
when entries are deleted at the same time. A customer reported a crash
and the backtrace showed thatctnetlink_dump_table was running while a
conntrack entry was being destroyed.
(see https://bugzilla.vyatta.com/show_bug.cgi?id=6402).

According to RCU documentation, when using hlist_nulls the reader
must handle the case of seeing a deleted entry and not proceed
further down the linked list.  The old code would continue
which caused the scan to walk into the free list.

This patch uses locking (rather than RCU) for this operation which
is guaranteed safe, and no longer requires getting reference while
doing dump operation.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

8 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux...
Linus Torvalds [Tue, 11 Jan 2011 22:45:52 +0000]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-udf-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-udf-2.6:
  UDF: Close small mem leak in udf_find_entry()
  udf: Fix directory corruption after extent merging
  udf: Protect udf_file_aio_write from possible races
  udf: Remove unnecessary bkl usages
  udf: Use of s_alloc_mutex to serialize udf_relocate_blocks() execution
  udf: Replace bkl with the UDF_I(inode)->i_data_sem for protect udf_inode_info struct
  udf: Remove BKL from free space counting functions
  udf: Call udf_add_free_space() for more blocks at once in udf_free_blocks()
  udf: Remove BKL from udf_put_super() and udf_remount_fs()
  udf: Protect default inode credentials by rwlock
  udf: Protect all modifications of LVID with s_alloc_mutex
  udf: Move handling of uniqueID into a helper function and protect it by a s_alloc_mutex
  udf: Remove BKL from udf_update_inode
  udf: Convert UDF_SB(sb)->s_flags to use bitops
  fs/udf: Add printf format/argument verification
  fs/udf: Use vzalloc

(Evil merge: this also removes the BKL dependency from the Kconfig file)

8 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Linus Torvalds [Tue, 11 Jan 2011 22:37:31 +0000]
Merge branch 'for_linus' of git://git./linux/kernel/git/tytso/ext4

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (44 commits)
  ext4: fix trimming starting with block 0 with small blocksize
  ext4: revert buggy trim overflow patch
  ext4: don't pass entire map to check_eofblocks_fl
  ext4: fix memory leak in ext4_free_branches
  ext4: remove ext4_mb_return_to_preallocation()
  ext4: flush the i_completed_io_list during ext4_truncate
  ext4: add error checking to calls to ext4_handle_dirty_metadata()
  ext4: fix trimming of a single group
  ext4: fix uninitialized variable in ext4_register_li_request
  ext4: dynamically allocate the jbd2_inode in ext4_inode_info as necessary
  ext4: drop i_state_flags on architectures with 64-bit longs
  ext4: reorder ext4_inode_info structure elements to remove unneeded padding
  ext4: drop ec_type from the ext4_ext_cache structure
  ext4: use ext4_lblk_t instead of sector_t for logical blocks
  ext4: replace i_delalloc_reserved_flag with EXT4_STATE_DELALLOC_RESERVED
  ext4: fix 32bit overflow in ext4_ext_find_goal()
  ext4: add more error checks to ext4_mkdir()
  ext4: ext4_ext_migrate should use NULL not 0
  ext4: Use ext4_error_file() to print the pathname to the corrupted inode
  ext4: use IS_ERR() to check for errors in ext4_error_file
  ...

8 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux...
Linus Torvalds [Tue, 11 Jan 2011 22:36:55 +0000]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs-2.6:
  ext2: Resolve 'dereferencing pointer to incomplete type' when enabling EXT2_XATTR_DEBUG
  ext3: Remove redundant unlikely()
  ext2: Remove redundant unlikely()
  ext3: speed up file creates by optimizing rec_len functions
  ext2: speed up file creates by optimizing rec_len functions
  ext3: Add more journal error check
  ext3: Add journal error check in resize.c
  quota: Use %pV and __attribute__((format (printf in __quota_error and fix fallout
  ext3: Add FITRIM handling
  ext3: Add batched discard support for ext3
  ext3: Add journal error check into ext3_rename()
  ext3: Use search_dirblock() in ext3_dx_find_entry()
  ext3: Avoid uninitialized memory references with a corrupted htree directory
  ext3: Return error code from generic_check_addressable
  ext3: Add journal error check into ext3_delete_entry()
  ext3: Add error check in ext3_mkdir()
  fs/ext3/super.c: Use printf extension %pV
  fs/ext2/super.c: Use printf extension %pV
  ext3: don't update sb journal_devnum when RO dev

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh...
Linus Torvalds [Tue, 11 Jan 2011 22:36:08 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/ericvh/v9fs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
  fs/9p: Don't set dentry->d_op in create routines
  fs/9p: fix spelling typo
  fs/9p: TREADLINK bugfix
  net/9p: Use proper data types
  fs/9p: Simplify the .L create operation
  fs/9p: Move dotl inode operations into a seperate file
  fs/9p: fix menu presentation
  fs/9p: Fix the return error on default acl removal
  fs/9p: Remove unnecessary semicolons

8 years agodrm/i915/debugfs: Show all objects in the gtt
Chris Wilson [Mon, 10 Jan 2011 00:00:24 +0000]
drm/i915/debugfs: Show all objects in the gtt

Useful for determining the layout.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Record AGP memory type upon error
Chris Wilson [Sun, 9 Jan 2011 21:07:49 +0000]
drm/i915: Record AGP memory type upon error

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Periodically flush the active lists and requests
Chris Wilson [Sun, 9 Jan 2011 21:05:44 +0000]
drm/i915: Periodically flush the active lists and requests

In order to retire active buffers whilst no client is active, we need to
insert our own flush requests onto the ring.

This is useful for servers that queue up some rendering and then go to
sleep as it allows us to the complete processing of those requests,
potentially making that memory available again much earlier.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agoah: reload pointers to skb data after calling skb_cow_data()
Dang Hongwu [Tue, 11 Jan 2011 07:13:33 +0000]
ah: reload pointers to skb data after calling skb_cow_data()

skb_cow_data() may allocate a new data buffer, so pointers on
skb should be set after this function.

Bug was introduced by commit dff3bb06 ("ah4: convert to ahash")
and 8631e9bd ("ah6: convert to ahash").

Signed-off-by: Wang Xuefu <xuefu.wang@6wind.com>
Acked-by: Krzysztof Witek <krzysztof.witek@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoah: update maximum truncated ICV length
Nicolas Dichtel [Tue, 11 Jan 2011 08:06:19 +0000]
ah: update maximum truncated ICV length

For SHA256, RFC4868 requires to truncate ICV length to 128 bits,
hence MAX_AH_AUTH_LEN should be updated to 16.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoxfrm: check trunc_len in XFRMA_ALG_AUTH_TRUNC
Nicolas Dichtel [Tue, 11 Jan 2011 08:04:12 +0000]
xfrm: check trunc_len in XFRMA_ALG_AUTH_TRUNC

Maximum trunc length is defined by MAX_AH_AUTH_LEN (in bytes)
and need to be checked when this value is set (in bits) by
the user. In ah4.c and ah6.c a BUG_ON() checks this condiftion.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoehea: Increase the skb array usage
Breno Leitao [Tue, 11 Jan 2011 07:45:57 +0000]
ehea: Increase the skb array usage

Currently the skb array is not fully allocated, and the allocation
is done as it's requested, which is not the expected way.

This patch just allocate the full skb array at driver initialization.
Also, this patch increases ehea version to 107.

Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agonet/fec: remove config FEC2 as it's used nowhere
Shawn Guo [Tue, 11 Jan 2011 02:07:52 +0000]
net/fec: remove config FEC2 as it's used nowhere

Signed-off-by: Shawn Guo <shawn.guo@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agopcnet_cs: add new_id
Ken Kawasaki [Tue, 11 Jan 2011 01:55:58 +0000]
pcnet_cs: add new_id

pcnet_cs:
     add another ID of "corega Ether CF-TD" 10Base-T PCMCIA card.

Signed-off-by: Ken Kawasaki <ken_kawasaki@spring.nifty.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agotcp: disallow bind() to reuse addr/port
Eric Dumazet [Tue, 11 Jan 2011 01:14:22 +0000]
tcp: disallow bind() to reuse addr/port

inet_csk_bind_conflict() logic currently disallows a bind() if
it finds a friend socket (a socket bound on same address/port)
satisfying a set of conditions :

1) Current (to be bound) socket doesnt have sk_reuse set
OR
2) other socket doesnt have sk_reuse set
OR
3) other socket is in LISTEN state

We should add the CLOSE state in the 3) condition, in order to avoid two
REUSEADDR sockets in CLOSE state with same local address/port, since
this can deny further operations.

Note : a prior patch tried to address the problem in a different (and
buggy) way. (commit fda48a0d7a8412ced tcp: bind() fix when many ports
are bound).

Reported-by: Gaspar Chilingarov <gasparch@gmail.com>
Reported-by: Daniel Baluta <daniel.baluta@gmail.com>
Tested-by: Daniel Baluta <daniel.baluta@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agodrm/i915/gtt: Unmap the PCI pages after unbinding them from the GTT
Chris Wilson [Tue, 11 Jan 2011 11:07:54 +0000]
drm/i915/gtt: Unmap the PCI pages after unbinding them from the GTT

Dave Airlie spotted that his ILK laptop with DMAR enabled was generating
the occasional DMAR warning.

"The ordering in the previous code was to rewrite the GTT table before
unmapping the pages and that makes sense to me."

This is his stable patch ported to d-i-n.

Reported-by: Dave Airlie <airlied@redhat.com>
Original-patch-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Record the error batchbuffer on each ring
Chris Wilson [Fri, 7 Jan 2011 21:06:07 +0000]
drm/i915: Record the error batchbuffer on each ring

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Include TLB miss overhead for computing WM
Chris Wilson [Sat, 8 Jan 2011 09:02:21 +0000]
drm/i915: Include TLB miss overhead for computing WM

The docs recommend that if 8 display lines fit inside the FIFO buffer,
then the number of watermark entries should be increased to hide the
latency of filling the rest of the FIFO buffer.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Propagate error from flushing the ring
Chris Wilson [Fri, 7 Jan 2011 17:09:48 +0000]
drm/i915: Propagate error from flushing the ring

... in order to avoid a BUG() and potential unbounded waits.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: detect & report PCH display error interrupts
Jesse Barnes [Tue, 4 Jan 2011 23:09:39 +0000]
drm/i915: detect & report PCH display error interrupts

FDI and the transcoders can fail for various reasons, so detect those
conditions and report on them.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: cleanup rc6 code
Jesse Barnes [Wed, 5 Jan 2011 20:01:26 +0000]
drm/i915: cleanup rc6 code

Cleanup several aspects of the rc6 code:
  - misnamed intel_disable_clock_gating function (was only about rc6)
  - remove commented call to intel_disable_clock_gating
  - rc6 enabling code belongs in its own function (allows us to move the
    actual clock gating enable call back into restore_state)
  - allocate power & render contexts up front, only free on unload
    (avoids ugly lazy init at rc6 enable time)

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[ickle: checkpatch cleanup]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: fix rc6 enabling around suspend/resume
Jesse Barnes [Wed, 5 Jan 2011 20:01:25 +0000]
drm/i915: fix rc6 enabling around suspend/resume

Enabling RC6 implies setting a graphics context.  Make sure we do that
only after the ring has been enabled, otherwise our ring commands will
hang.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: re-enable rc6 support for Ironlake+
Jesse Barnes [Wed, 5 Jan 2011 20:01:24 +0000]
drm/i915: re-enable rc6 support for Ironlake+

Re-enable rc6 support on Ironlake for power savings.  Adds a debugfs
file to check current RC state, adds a missing workaround for Ironlake
MI_SET_CONTEXT instructions, and renames MCHBAR_RENDER_STANDBY to
RSTDBYCTL to match the docs.

Keep RC6 and the power context disabled on pre-ILK.  It only seems to
hang and doesn't save any power.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Make the ring IMR handling private
Chris Wilson [Wed, 5 Jan 2011 10:32:24 +0000]
drm/i915: Make the ring IMR handling private

As the IMR for the USER interrupts are not modified elsewhere, we can
separate the spinlock used for these from that of hpd and pipestats.
Those two IMR are manipulated under an IRQ and so need heavier locking.

Reported-and-tested-by: Alexey Fisher <bug-track@fisher-privat.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915/ringbuffer: Simplify the ring irq refcounting
Chris Wilson [Tue, 4 Jan 2011 22:22:56 +0000]
drm/i915/ringbuffer: Simplify the ring irq refcounting

... and move it under the spinlock to gain the appropriate memory
barriers.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=32752
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915/debugfs: Show the per-ring IMR
Chris Wilson [Tue, 4 Jan 2011 22:22:17 +0000]
drm/i915/debugfs: Show the per-ring IMR

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Mask USER interrupts on gen6 (until required)
Chris Wilson [Tue, 4 Jan 2011 17:35:21 +0000]
drm/i915: Mask USER interrupts on gen6 (until required)

Otherwise we may consume 20% of the CPU just handling IRQs whilst
rendering. Ouch.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Handle ringbuffer stalls when flushing
Chris Wilson [Tue, 4 Jan 2011 17:34:02 +0000]
drm/i915: Handle ringbuffer stalls when flushing

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Enforce write ordering through the GTT
Chris Wilson [Tue, 4 Jan 2011 18:42:07 +0000]
drm/i915: Enforce write ordering through the GTT

We need to ensure that writes through the GTT land before any
modification to the MMIO registers and so must impose a mandatory write
barrier when flushing the GTT domain. This was revealed by relaxing the
write ordering by experimentally mapping the registers and the GATT as
write-combining.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Remove impossible test
Chris Wilson [Mon, 3 Jan 2011 16:39:21 +0000]
drm/i915: Remove impossible test

As has_gem is unconditionally set to true, the conditional immediately
following that assignment is superfluous.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: avoid reading non-existent PLL reg on Ironlake+
Jesse Barnes [Thu, 30 Dec 2010 17:36:39 +0000]
drm/i915: avoid reading non-existent PLL reg on Ironlake+

These functions need to be reworked for Ironlake and above, but until
then at least avoid reading non-existent registers.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[ickle: combine with a gratuitous tidy]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: add 'reset' parameter
Chris Wilson [Thu, 23 Dec 2010 13:33:15 +0000]
drm/i915: add 'reset' parameter

When bringing up new hardware, or otherwise experimenting, GPU hangs are
a way of life. However, the automatic GPU reset can do more harm than
good under these circumstances, as we may wish to capture a full trace for
debugging.

Based on a patch by Zhenyu Wang.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: fix the wrong latency value while computing wm0
Yuanhan Liu [Thu, 23 Dec 2010 08:35:40 +0000]
drm/i915: fix the wrong latency value while computing wm0

On Ironlake, the LP0 latency is hardcoded and in ns unit, while on
Sandybridge, it comes from a register and with unit 0.1 us. So, fix
the wrong latency value while computing wm0 on Ironlake and Sandybridge.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: support low power watermarks on Ironlake
Jesse Barnes [Tue, 21 Dec 2010 21:10:23 +0000]
drm/i915: support low power watermarks on Ironlake

This patch actually makes the watermark code even uglier (if that's
possible), but has the advantage of sharing code between SNB and ILK at
least.  Longer term we should refactor the watermark stuff into its own
file and clean it up now that we know how it's supposed to work.

Supporting WM2 on my Vaio reduced power consumption by around 0.5W, so
this patch is definitely worthwhile (though it also needs lots of test
coverage).

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[ickle: pass the watermark structs arounds]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agoagp/intel: Flush the chipset write buffers when changing GTT base
Chris Wilson [Thu, 23 Dec 2010 10:40:38 +0000]
agp/intel: Flush the chipset write buffers when changing GTT base

Flush the chipset write buffers before and after adjusting the GTT base
register, just in case. We only modify this value upon initialisation
(boot and resume) so there should be no outstanding writes, however
there are always those persistent PGTBL_ER that keep getting reported
upon resume.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Workaround erratum on i830 for TAIL pointer within last 2 cachelines
Chris Wilson [Wed, 22 Dec 2010 14:04:47 +0000]
drm/i915: Workaround erratum on i830 for TAIL pointer within last 2 cachelines

On i830 if the tail pointer is set to within 2 cachelines of the end of
the buffer, the chip may hang. So instead if the tail were to land in
that location, we pad the end of the buffer with NOPs, and start again
at the beginning.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: Use the mappable sizes determined by GTT for consistency.
Chris Wilson [Wed, 22 Dec 2010 14:07:12 +0000]
drm/i915: Use the mappable sizes determined by GTT for consistency.

There should be no difference, but we can eliminate redundant code.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: support overclocking on Sandy Bridge
Jesse Barnes [Mon, 20 Dec 2010 19:34:20 +0000]
drm/i915: support overclocking on Sandy Bridge

In some configuration, the PCU may allow us to overclock the GPU.
Check for this case and adjust the max frequency as appropriate.  Also
initialize the min/max frequencies to default values as indicated by
hardware.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915/lvds: Always use 0 to disable the pfit controller
Chris Wilson [Tue, 11 Jan 2011 18:09:58 +0000]
drm/i915/lvds: Always use 0 to disable the pfit controller

... and just any combination of bits & ~PFIT_ENABLE. This way we do not
attempt disable to the panel fitter controller uselessly upon
intel_lvds_disable().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915/panel: Only record the backlight level when it is enabled
Chris Wilson [Tue, 11 Jan 2011 17:06:04 +0000]
drm/i915/panel: Only record the backlight level when it is enabled

By tracking the current status of the backlight we can prevent recording
the value of the current backlight when we have disabled it. And so
prevent restoring it to 'off' after an unbalanced sequence of
intel_lvds_disable/enable.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=22672
Tested-by: Alex Riesen <raa.lkml@gmail.com>
Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

8 years agodrm/i915/sdvo: Defer detection of output capabilities until probing
Chris Wilson [Tue, 4 Jan 2011 20:10:52 +0000]
drm/i915/sdvo: Defer detection of output capabilities until probing

Alex Fiestas reported an issue with his HDMI connector being misdetected
as DVI unless he had something connected upon boot. By moving the
decision as to whether to use HDMI or DVI encoding for the HDMI capable
output until we probe the monitor means that we should avoid sending a
HDMI signal to a DVI monitor and also correctly detect hardware like
Alex's.

However, to really determine what connector is soldered onto the wire we
need to inspect the VBT sdvo child devices - but can we trust it?

Reported-by: Alex Fiestas <alex@eyeos.org>
Tested-by: Alex Fiestas <alex@eyeos.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=32828
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

8 years agodrm/i915: fix calculation of eDP signal levels on Sandybridge
Yuanhan Liu [Thu, 6 Jan 2011 10:26:08 +0000]
drm/i915: fix calculation of eDP signal levels on Sandybridge

Some voltage swing/pre-emphasis level use the same value on eDP
Sandybridge, like 400mv_0db and 600mv_0db are with the same value
of (0x0 << 22). So, fix them, and point out the value if it isn't
a supported voltage swing/pre-emphasis level.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

8 years agodrm/i915/crt: Check for a analog monitor in case of DVI-I
David Müller [Thu, 6 Jan 2011 12:29:32 +0000]
drm/i915/crt: Check for a analog monitor in case of DVI-I

Since Linux 2.6.36 the digital output on my system (855GME + DVI-I) is
not working any longer. The analog output is always activated
regardless of the type of monitor attached.

The culprit seems to be intel_crt_detect_ddc(), which returns true as
soon as an ACK from the EDID device is received. Obviously this
approach does not work with DVI-I where the analog and digital outputs
share a common DDC bus.

In a similar manner to the shared DDC wire, ala the "Mac Mini Hack", we
need an additional check to make sure that there really is an analog
device attached to the DDC.

Signed-off-by: David Müller <d.mueller@elsoft.ch>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

8 years agodrm/i915: make DP training try a little harder
Jesse Barnes [Wed, 5 Jan 2011 22:45:24 +0000]
drm/i915: make DP training try a little harder

When trying to do channel equalization, we need to make sure we still
have clock recovery on all lanes while training.  We also need to try
clock recovery again if we lose the clock or if channel eq fails 5
times.  We'll try clock recovery up to 5 more times before giving up
entirely.

Gets suspend/resume working on my Vaio again and brings us back into
compliance with the DP training sequence spec.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

8 years agodrm/i915: check eDP encoder correctly when setting modes
Jesse Barnes [Tue, 4 Jan 2011 18:46:49 +0000]
drm/i915: check eDP encoder correctly when setting modes

We were using a stale pointer in the check which caused us to use CPU
attached DP params when we should have been using PCH attached params.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=31988
Tested-by: Jan-Hendrik Zab <jan@jhz.name>
Tested-by: Christoph Lukas <christoph.lukas@gmx.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

8 years agoext4: fix trimming starting with block 0 with small blocksize
Jan Kara [Tue, 11 Jan 2011 20:16:31 +0000]
ext4: fix trimming starting with block 0 with small blocksize

When s_first_data_block is not zero (which happens e.g. when block size is 1KB)
and trim ioctl is called to start trimming from block 0, the math in
ext4_get_group_no_and_offset() overflows. The overall result is that ioctl
returns EINVAL which is kind of unexpected and we probably don't want
userspace tools to bother with internal details of filesystem structure.
So just silently increase starting offset (and shorten length) when starting
block is below s_first_data_block.

CC: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: revert buggy trim overflow patch
Theodore Ts'o [Tue, 11 Jan 2011 19:42:29 +0000]
ext4: revert buggy trim overflow patch

This reverts commit 4f531501e44: ext4: fix possible overflow in
ext4_trim_fs()

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoMerge branch 'for-linus-merged' of git://oss.sgi.com/xfs/xfs
Linus Torvalds [Tue, 11 Jan 2011 19:42:06 +0000]
Merge branch 'for-linus-merged' of git://oss.sgi.com/xfs/xfs

* 'for-linus-merged' of git://oss.sgi.com/xfs/xfs: (47 commits)
  xfs: convert grant head manipulations to lockless algorithm
  xfs: introduce new locks for the log grant ticket wait queues
  xfs: convert log grant heads to atomic variables
  xfs: convert l_tail_lsn to an atomic variable.
  xfs: convert l_last_sync_lsn to an atomic variable
  xfs: make AIL tail pushing independent of the grant lock
  xfs: use wait queues directly for the log wait queues
  xfs: combine grant heads into a single 64 bit integer
  xfs: rework log grant space calculations
  xfs: fact out common grant head/log tail verification code
  xfs: convert log grant ticket queues to list heads
  xfs: use AIL bulk delete function to implement single delete
  xfs: use AIL bulk update function to implement single updates
  xfs: remove all the inodes on a buffer from the AIL in bulk
  xfs: consume iodone callback items on buffers as they are processed
  xfs: reduce the number of AIL push wakeups
  xfs: bulk AIL insertion during transaction commit
  xfs: clean up xfs_ail_delete()
  xfs: Pull EFI/EFD handling out from under the AIL lock
  xfs: fix EFI transaction cancellation.
  ...

8 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec...
Linus Torvalds [Tue, 11 Jan 2011 19:28:34 +0000]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jlbec/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2: (22 commits)
  MAINTAINERS: Update Joel Becker's email address
  ocfs2: Remove unused truncate function from alloc.c
  ocfs2/cluster: dereferencing before checking in nst_seq_show()
  ocfs2: fix build for OCFS2_FS_STATS not enabled
  ocfs2/cluster: Show o2net timing statistics
  ocfs2/cluster: Track process message timing stats for each socket
  ocfs2/cluster: Track send message timing stats for each socket
  ocfs2/cluster: Use ktime instead of timeval in struct o2net_sock_container
  ocfs2/cluster: Replace timeval with ktime in struct o2net_send_tracking
  ocfs2: Add DEBUG_FS dependency
  ocfs2/dlm: Hard code the values for enums
  ocfs2/dlm: Minor cleanup
  ocfs2/dlm: Cleanup dlmdebug.c
  ocfs2: Release buffer_head in case of error in ocfs2_double_lock.
  ocfs2/cluster: Pin the local node when o2hb thread starts
  ocfs2/cluster: Show pin state for each o2hb region
  ocfs2/cluster: Pin/unpin o2hb regions
  ocfs2/cluster: Remove dropped region from o2hb quorum region bitmap
  ocfs2/cluster: Pin the remote node item in configfs
  ocfs2/dlm: make existing convertion precedent over new lock
  ...

8 years agoNFS fix the setting of exchange id flag
Andy Adamson [Tue, 14 Dec 2010 15:11:57 +0000]
NFS fix the setting of exchange id flag

Indicate support for referrals. Do not set any PNFS roles. Check the flags
returned by the server for validity. Do not use exchange flags from an old
client ID instance when recovering a client ID.

Update the EXCHID4_FLAG_XXX set to RFC 5661.

Signed-off-by: Andy Adamson <andros@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

8 years agoMerge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze
Linus Torvalds [Tue, 11 Jan 2011 19:13:46 +0000]
Merge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze

* 'next' of git://git.monstr.eu/linux-2.6-microblaze:
  microblaze: remove obsolete DEBUG_BOOTMEM
  microblaze: trivial: Fix removed the part of line
  microblaze: Fix __muldi3 function for little-endian.
  microblaze: Clear i/dcache for sw breakpoints
  microblaze: Remove useless early_init_dt_check_for_initrd
  microblaze: Fix unaligned exception for little endian platform
  microblaze: Add PVR for Microblaze v8.00.b
  microblaze: Correct PVR access macros
  Revert "microblaze: Simplify syscall rutine"
  microblaze: Fix initramfs
  arch/microblaze: Remove unnecessary semicolons

8 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 11 Jan 2011 19:11:46 +0000]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Fix Moorestown VRTC fixmap placement
  x86/gpio: Implement x86 gpio_to_irq convert function
  x86, UV: Fix APICID shift for Westmere processors
  x86: Use PCI method for enabling AMD extended config space before MSR method
  x86: tsc: Prevent delayed init if initial tsc calibration failed
  x86, lapic-timer: Increase the max_delta to 31 bits
  x86: Fix sparse non-ANSI function warnings in smpboot.c
  x86, numa: Fix CONFIG_DEBUG_PER_CPU_MAPS without NUMA emulation
  x86, AMD, PCI: Add AMD northbridge PCI device id for CPU families 12h and 14h
  x86, numa: Fix cpu to node mapping for sparse node ids
  x86, numa: Fake node-to-cpumask for NUMA emulation
  x86, numa: Fake apicid and pxm mappings for NUMA emulation
  x86, numa: Avoid compiling NUMA emulation functions without CONFIG_NUMA_EMU
  x86, numa: Reduce minimum fake node size to 32M

Fix up trivial conflict in arch/x86/kernel/apic/x2apic_uv_x.c

8 years agoMerge branch 'timers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 11 Jan 2011 19:06:41 +0000]
Merge branch 'timers-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'timers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  rtc: Namespace fixup
  RTC: Remove UIE emulation
  RTC: Rework RTC code to use timerqueue for events

Fix up trivial conflict in drivers/rtc/rtc-dev.c

8 years agoMerge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 11 Jan 2011 19:02:13 +0000]
Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (28 commits)
  perf session: Fix infinite loop in __perf_session__process_events
  perf evsel: Support perf_evsel__open(cpus > 1 && threads > 1)
  perf sched: Use PTHREAD_STACK_MIN to avoid pthread_attr_setstacksize() fail
  perf tools: Emit clearer message for sys_perf_event_open ENOENT return
  perf stat: better error message for unsupported events
  perf sched: Fix allocation result check
  perf, x86: P4 PMU - Fix unflagged overflows handling
  dynamic debug: Fix build issue with older gcc
  tracing: Fix TRACE_EVENT power tracepoint creation
  tracing: Fix preempt count leak
  tracepoint: Add __rcu annotation
  tracing: remove duplicate null-pointer check in skb tracepoint
  tracing/trivial: Add missing comma in TRACE_EVENT comment
  tracing: Include module.h in define_trace.h
  x86: Save rbp in pt_regs on irq entry
  x86, dumpstack: Fix unused variable warning
  x86, NMI: Clean-up default_do_nmi()
  x86, NMI: Allow NMI reason io port (0x61) to be processed on any CPU
  x86, NMI: Remove DIE_NMI_IPI
  x86, NMI: Add priorities to handlers
  ...

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc
Linus Torvalds [Tue, 11 Jan 2011 19:01:24 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/cjb/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc: (39 commits)
  mmc: davinci: add support for SDIO irq handling
  mmc: fix division by zero in MMC core
  mmc: tmio_mmc: fix CMD irq handling
  mmc: tmio_mmc: handle missing HW interrupts
  mfd: sh_mobile_sdhi: activate SDIO IRQ for tmio_mmc
  mmc: tmio_mmc: implement SDIO IRQ support
  mfd: sdhi: require the tmio-mmc driver to bounce unaligned buffers
  mmc: tmio_mmc: silence compiler warnings
  mmc: tmio_mmc: implement a bounce buffer for unaligned DMA
  mmc: tmio_mmc: merge the private header into the driver
  mmc: tmio_mmc: fix PIO fallback on DMA descriptor allocation failure
  mmc: tmio_mmc: allow multi-element scatter-gather lists
  mmc: Register debugfs dir before calling card probe function.
  mmc: MMC_BLOCK_MINORS should depend on MMC_BLOCK.
  mmc: Explain why we make adjacent mmc_bus_{put,get} calls during rescan.
  mmc: Fix sd/sdio/mmc initialization frequency retries
  mmc: fix mmc_set_bus_width_ddr() call without bus-width-test cap
  mmc: dw_mmc: Add Synopsys DesignWare mmc host driver.
  mmc: add sdhci-tegra driver for Tegra SoCs
  mmc: sdhci: add quirk for max len ADMA descriptors
  ...

8 years agofs/9p: Don't set dentry->d_op in create routines
Aneesh Kumar K.V [Tue, 11 Jan 2011 08:14:47 +0000]
fs/9p: Don't set dentry->d_op in create routines

We do set dentry->d_op in lookup even in case of EOENT entries.
That implies we should have dentry->d_op already set when
create/mkdir/mknod/link/symlink routines are called

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agofs/9p: fix spelling typo
Eric Van Hensbergen [Tue, 11 Jan 2011 15:49:03 +0000]
fs/9p: fix spelling typo

introduced a typo somehow during a hand merge

Reported by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agofs/9p: TREADLINK bugfix
M. Mohan Kumar [Sat, 8 Jan 2011 01:58:46 +0000]
fs/9p: TREADLINK bugfix

Remove v9fs_vfs_readlink_dotl function and use generic_readlink. Update
v9fs_vfs_follow_link_dotl function to accommodate this change

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Reported-by:  Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agonet/9p: Use proper data types
M. Mohan Kumar [Mon, 10 Jan 2011 20:23:53 +0000]
net/9p: Use proper data types

Use proper data types for storing the count of the binary blob and
length of a string. Without this patch length calculation of string will
always result in -1 because of comparision between signed and unsigned
integer.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agofs/9p: Simplify the .L create operation
Aneesh Kumar K.V [Mon, 10 Jan 2011 20:22:21 +0000]
fs/9p: Simplify the .L create operation

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agofs/9p: Move dotl inode operations into a seperate file
Aneesh Kumar K.V [Mon, 10 Jan 2011 19:51:47 +0000]
fs/9p: Move dotl inode operations into a seperate file

Source Code Reorganization

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agofs/9p: fix menu presentation
Randy Dunlap [Sun, 2 Jan 2011 06:15:54 +0000]
fs/9p: fix menu presentation

Make the 9P_FS kconfig options subordinate to the 9P_FS kconfig symbol
in the menu presentation instead of them all being at the same level.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agofs/9p: Fix the return error on default acl removal
Aneesh Kumar K.V [Fri, 10 Dec 2010 06:49:31 +0000]
fs/9p: Fix the return error on default acl removal

If we don't have default ACL, then trying to remove
default acl on a file should return 0.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>

8 years agofs/9p: Remove unnecessary semicolons
Joe Perches [Mon, 15 Nov 2010 03:04:51 +0000]
fs/9p: Remove unnecessary semicolons

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

8 years agox86: Fix Moorestown VRTC fixmap placement
Arjan van de Ven [Tue, 11 Jan 2011 10:55:56 +0000]
x86: Fix Moorestown VRTC fixmap placement

The x86 fixmaps need to be all together... unfortunately the
VRTC one was misplaced.

This patch makes sure the MRST VRTC fixmap is put prior to the
__end_of_permanent_fixed_addresses marker.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
LKML-Reference: <20110111105544.24448.27607.stgit@bob.linux.org.uk>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agox86/gpio: Implement x86 gpio_to_irq convert function
Alek Du [Tue, 11 Jan 2011 10:55:32 +0000]
x86/gpio: Implement x86 gpio_to_irq convert function

We need this for x86 MID platforms where GPIO interrupts are
used. No special magic is needed so the default 1:1 behaviour
will do nicely.

Signed-off-by: Alek Du <alek.du@intel.com>
Signed-off-by: Jacob Pan <jacob.jun.pan@intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
LKML-Reference: <20110111105439.24448.69863.stgit@bob.linux.org.uk>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agox86, UV: Fix APICID shift for Westmere processors
Jack Steiner [Mon, 10 Jan 2011 19:52:11 +0000]
x86, UV: Fix APICID shift for Westmere processors

Westmere processors use a different algorithm for
assigning APICIDs on SGI UV systems. The location of the
node number within the apicid is now a function of the
processor type.

Signed-off-by: Jack Steiner <steiner@sgi.com>
LKML-Reference: <20110110195210.GA18737@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agox86: Use PCI method for enabling AMD extended config space before MSR method
Jan Beulich [Mon, 10 Jan 2011 16:20:23 +0000]
x86: Use PCI method for enabling AMD extended config space before MSR method

While both methods should work equivalently well for the native
case, the Xen Dom0 case can't reliably work with the MSR one,
since there's no guarantee that the virtual CPUs it has
available fully cover all necessary physical ones.

As per the suggestion of Robert Richter the patch only adds the
PCI method, but leaves the MSR one as a fallback to cover new
systems the PCI IDs of which may not have got added to the code
base yet.

The only change in v2 is the breaking out of the new CPI
initialization method into a separate function, as requested by
Ingo.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Acked-by: Robert Richter <robert.richter@amd.com>
Cc: Andreas Herrmann3 <Andreas.Herrmann3@amd.com>
Cc: Joerg Roedel <joerg.roedel@amd.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
LKML-Reference: <4D2B3FD7020000780002B67D@vpn.id2.novell.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agox86: tsc: Prevent delayed init if initial tsc calibration failed
Thomas Gleixner [Tue, 11 Jan 2011 10:40:48 +0000]
x86: tsc: Prevent delayed init if initial tsc calibration failed

commit a8760ec (x86: Check tsc available/disabled in the delayed init
function) missed to prevent the setup of the delayed init function in
case the initial tsc calibration failed. This results in the same
divide by zero bug as we have seen without the tsc disabled check.

Skip the delayed work setup when tsc_khz (the initial calibration
value) is 0.

Bisected-and-tested-by: Kirill A. Shutemov <kas@openvz.org>
Cc: John Stultz <john.stultz@linaro.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

8 years agopowerpc/pseries: Fix VPHN build errors on non-SMP systems
Jesse Larrew [Fri, 17 Dec 2010 22:07:47 +0000]
powerpc/pseries: Fix VPHN build errors on non-SMP systems

The header asm/hvcall.h was previously included indirectly via
smp.h. On non-SMP systems, however, these declarations are excluded
and the build breaks. This is easily fixed by including asm/hvcall.h
directly.

The VPHN feature is only meaningful on NUMA systems that implement
the SPLPAR option, so exclude the VPHN code on systems without
SPLPAR enabled.

Also, expose unmap_cpu_from_node() on systems with SPLPAR enabled,
even if CONFIG_HOTPLUG_CPU is disabled.

Lastly, map_cpu_to_node() is now needed by VPHN to manipulate the
node masks after boot time, so remove the __cpuinit annotation to
fix a section mismatch.

Signed-off-by: Jesse Larrew <jlarrew@linux.vnet.ibm.com>

8 years agoMerge remote branch 'gcl/next' into next
Benjamin Herrenschmidt [Tue, 11 Jan 2011 04:10:08 +0000]
Merge remote branch 'gcl/next' into next

8 years agoMerge branch 'master' into for-linus-merged
Alex Elder [Tue, 11 Jan 2011 03:35:55 +0000]
Merge branch 'master' into for-linus-merged

This merge pulls the XFS master branch into the latest Linus master.
This results in a merge conflict whose best fix is not obvious.
I manually fixed the conflict, in "fs/xfs/xfs_iget.c".

Dave Chinner had done work that resulted in RCU freeing of inodes
separate from what Nick Piggin had done, and their results differed
slightly in xfs_inode_free().  The fix updates Nick's call_rcu()
with the use of VFS_I(), while incorporating needed updates to some
XFS inode fields implemented in Dave's series.  Dave's RCU callback
function has also been removed.

Signed-off-by: Alex Elder <aelder@sgi.com>

8 years agoMerge branches 'cxgb4', 'ipath', 'ipoib', 'mlx4', 'mthca', 'nes', 'qib' and 'srp...
Roland Dreier [Tue, 11 Jan 2011 01:43:30 +0000]
Merge branches 'cxgb4', 'ipath', 'ipoib', 'mlx4', 'mthca', 'nes', 'qib' and 'srp' into for-next

8 years agoIB/qib: Fix refcount leak in lkey/rkey validation
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:23 +0000]
IB/qib: Fix refcount leak in lkey/rkey validation

The mr optimization introduced a reference count leak on an exception
test.  The lock/refcount manipulation is moved down and the problematic
exception test now calls bail to insure that the lock is released.

Additional fixes as suggested by Ralph Campbell <ralph.campbell@qlogic.org>:
- reduce lock scope of dma regions
- use explicit values on returns vs. automatic ret value

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Improve SERDES tunning on QMH boards
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:22 +0000]
IB/qib: Improve SERDES tunning on QMH boards

Improve the QMH SERDES tunning on initial driver load by having the
driver go through a link state change.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Unnecessary delayed completions on RC connection
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:22 +0000]
IB/qib: Unnecessary delayed completions on RC connection

Currently on receipt of a response message (ACKs, RDMA Response,
Atomic Responses etc.) if the SDMA completion counter is not advanced
the driver delays the completion of the WQE.  In most cases this is
overly pessimistic as the response (ACK) to a previously transmitted
send implies that the send is complete.  Ensure that SDMA queue is
progressed appropriately before determining if a send has delayed
completions.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Issue pre-emptive NAKs on eager buffer overflow
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:22 +0000]
IB/qib: Issue pre-emptive NAKs on eager buffer overflow

Under congestion resulting in eager buffer overflow attempt to send
pre-emptive NAKs if header queue entries with TID errors are generated
and a valid header is present.  This prevents long timeouts and flow
restarts if a trailing set of packets are dropped due to eager
overflows.  Pre-emptive NAKs are currently only supported for RDMA
writes.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: RDMA lkey/rkey validation is inefficient for large MRs
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:22 +0000]
IB/qib: RDMA lkey/rkey validation is inefficient for large MRs

The current code loops during rkey/lkey validiation to isolate the MR
for the RDMA, which is expensive when the current operation is inside
a very large memory region.

This fix optimizes rkey/lkey validation routines for user memory
regions and fast memory regions.  The MR entry can be isolated by
shifts/mods instead of looping.  The existing loop is preserved for
phys memory regions for now.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Change QPN increment
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:22 +0000]
IB/qib: Change QPN increment

Changing from +1 to +2 allows for better QP distribution across
receive contexts.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Add fix missing from earlier patch
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:21 +0000]
IB/qib: Add fix missing from earlier patch

The upstream code was missing part of a receive/error race fix from
the internal tree.  Add the missing part, which makes future merges
possible.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Change receive queue/QPN selection
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:21 +0000]
IB/qib: Change receive queue/QPN selection

The basic idea is that on SusieQ, the difficult part of mapping QPN to
context is handled by the mapping registers so the generic QPN
allocation doesn't need to worry about chip specifics.  For Monty and
Linda, there is no mapping table so the qpt->mask (same as
dd->qpn_mask), is used to see if the QPN to context falls within
[zero..dd->n_krcv_queues).

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Fix interrupt mitigation
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:21 +0000]
IB/qib: Fix interrupt mitigation

For SusieQ we need to write to the interrupt timer register before
updating the header queue head with interrupt count.  This is to
ensure that the timer is enabled properly and a receive available
interrupt is delivered.  Otherwise this interrupt can be lost if the
receiver header/eager queues are full before the timer is enabled.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Avoid duplicate writes to the rcv head register
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:21 +0000]
IB/qib: Avoid duplicate writes to the rcv head register

Avoid duplicate writes to the head register as this can lead to lost
interrupts if the context goes full before the second write is done.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Add a few new SERDES tunings
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:21 +0000]
IB/qib: Add a few new SERDES tunings

Add new SERDES tuning to aid manufacturing.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

8 years agoIB/qib: Reset packet list after freeing
Mike Marciniszyn [Tue, 11 Jan 2011 01:42:21 +0000]
IB/qib: Reset packet list after freeing

Reset the list pointers after freeing the SDMA packet list.  This is
done to any potential double-free cases.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>