Commit graph

254,703 commits

Author SHA1 Message Date
Alexander Clouter
e25bac968d [ARM] orion5x: accelerate NAND on the TS-78xx
The NAND supports 32bit reads and writes so lets stop shunting 8bit
chunks across the bus.

Doing a dumb 'dd' benchmark, this increases performance roughly like so:
 * read: 1.3MB/s to 3.4MB/s
 * write: 614kB/s to 882kB/s

Signed-off-by: Alexander Clouter <alex@digriz.org.uk>
Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
2011-03-03 15:53:34 -05:00
Linus Torvalds
4c7fd114c6 Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs
* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: zero proper structure size for geometry calls
2011-03-03 12:44:22 -08:00
Linus Torvalds
c640e13f8e Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2:
  nilfs2: fix regression that i-flag is not set on changeless checkpoints
2011-03-03 12:42:48 -08:00
Joe Perches
6b8a66ee91 tun: Convert logging messages to pr_<level> and tun_debug
Use the current logging forms with pr_fmt.
Convert DBG macro to tun_debug, use netdev_printk as well.
Add printf verification when TUN_DEBUG not defined.
Miscellaneous comment typo fix.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 12:21:14 -08:00
Harry Ciao
2ad18bdf3b SELinux: Compute SID for the newly created socket
The security context for the newly created socket shares the same
user, role and MLS attribute as its creator but may have a different
type, which could be specified by a type_transition rule in the relevant
policy package.

Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
[fix call to security_transition_sid to include qstr, Eric Paris]
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Stephen Smalley <sds@tycho.nsa.gov>
2011-03-03 15:19:44 -05:00
Harry Ciao
6f5317e730 SELinux: Socket retains creator role and MLS attribute
The socket SID would be computed on creation and no longer inherit
its creator's SID by default. Socket may have a different type but
needs to retain the creator's role and MLS attribute in order not
to break labeled networking and network access control.

The kernel value for a class would be used to determine if the class
if one of socket classes. If security_compute_sid is called from
userspace the policy value for a class would be mapped to the relevant
kernel value first.

Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Stephen Smalley <sds@tycho.nsa.gov>
2011-03-03 15:19:43 -05:00
Harry Ciao
4bc6c2d5d8 SELinux: Auto-generate security_is_socket_class
The security_is_socket_class() is auto-generated by genheaders based
on classmap.h to reduce maintenance effort when a new class is defined
in SELinux kernel. The name for any socket class should be suffixed by
"socket" and doesn't contain more than one substr of "socket".

Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Stephen Smalley <sds@tycho.nsa.gov>
2011-03-03 15:19:43 -05:00
Po-Yu Chuang
8d77c036b5 net: add Faraday FTMAC100 10/100 Ethernet driver
FTMAC100 Ethernet Media Access Controller supports 10/100 Mbps and
MII.  This driver has been working on some ARM/NDS32 SoC's including
Faraday A320 and Andes AG101.

Signed-off-by: Po-Yu Chuang <ratbert@faraday-tech.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 12:19:11 -08:00
David S. Miller
29546a6404 ipv6: Use ERR_CAST in addrconf_dst_alloc.
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 12:10:37 -08:00
Stanislaw Gruszka
ba04c7c93b r8169: disable ASPM
For some time is known that ASPM is causing troubles on r8169, i.e. make
device randomly stop working without any errors in dmesg.

Currently Tomi Leppikangas reports that system with r8169 device hangs
with MCE errors when ASPM is enabled:
https://bugzilla.redhat.com/show_bug.cgi?id=642861#c4

Lets disable ASPM for r8169 devices at all, to avoid problems with
r8169 PCIe devices at least for some users.

Reported-by: Tomi Leppikangas <tomi.leppikangas@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 11:55:43 -08:00
David S. Miller
a45d49d105 Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/jkirsher/net-next-2.6 2011-03-03 11:28:35 -08:00
David S. Miller
e066008b38 ipv4: Fix __ip_dev_find() to use ifa_local instead of ifa_address.
Reported-by: Stephen Hemminger <shemminger@vyatta.com>
Reported-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 11:24:19 -08:00
Eric Dumazet
d276055c4e net_sched: reduce fifo qdisc size
Because of various alignements [SLUB / qdisc], we use 512 bytes of
memory for one {p|b}fifo qdisc, instead of 256 bytes on 64bit arches and
192 bytes on 32bit ones.

Move the "u32 limit" inside "struct Qdisc" (no impact on other qdiscs)

Change qdisc_alloc(), first trying a regular allocation before an
oversized one.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 11:10:02 -08:00
Patrick McHardy
c53fa1ed92 netlink: kill loginuid/sessionid/sid members from struct netlink_skb_parms
Netlink message processing in the kernel is synchronous these days, the
session information can be collected when needed.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 10:55:40 -08:00
David S. Miller
06dc94b1ed ipv4: Fix crash in dst_release when udp_sendmsg route lookup fails.
As reported by Eric:

[11483.697233] IP: [<c12b0638>] dst_release+0x18/0x60
 ...
[11483.697741] Call Trace:
[11483.697764]  [<c12fc9d2>] udp_sendmsg+0x282/0x6e0
[11483.697790]  [<c12a1c01>] ? memcpy_toiovec+0x51/0x70
[11483.697818]  [<c12dbd90>] ? ip_generic_getfrag+0x0/0xb0

The pointer passed to dst_release() is -EINVAL, that's because
we leave an error pointer in the local variable "rt" by accident.

NULL it out to fix the bug.

Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-03-03 10:38:01 -08:00
Tony Lindgren
077f8ec889 Merge branch 'for_2.6.38/pm-fixes' of ssh://master.kernel.org/pub/scm/linux/kernel/git/khilman/linux-omap-pm into devel-fixes 2011-03-03 10:25:18 -08:00
Hari Kanigeri
525a11381b omap: mailbox: resolve hang issue
omap4 interrupt disable bits is different. On rx kfifo full, the mbox rx
interrupts wasn't getting disabled, and this is causing the rcm stress tests
to hang.

Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>
Signed-off-by: Armando Uribe <x0095078@ti.com>
Signed-off-by: Fernando Guzman Lugo <fernando.lugo@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
2011-03-03 10:24:19 -08:00
Sage Weil
16a8b70a5a ceph: do not clear I_COMPLETE from d_release
First, this was racy anyway: d_release isn't called until well after the
dentry is unhashed.  Second, this runs afoul of the recent dcache change
that clears d_parent prior to calling d_release (949854d0), causing a NULL
pointer dereference.

Signed-off-by: Sage Weil <sage@newdream.net>
2011-03-03 10:09:52 -08:00
Sage Weil
b545cc1505 ceph: do not set I_COMPLETE
Do not set the I_COMPLETE flag on directories until we resolve races with
dcache pruning.

Signed-off-by: Sage Weil <sage@newdream.net>
2011-03-03 10:09:51 -08:00
Sage Weil
9bde178d05 Revert "ceph: keep reference to parent inode on ceph_dentry"
This reverts commit 97d79b403e.

This fails to account for d_parent changes due to rename or disconnected
dentries due to submounts or NFS reexports.

Signed-off-by: Sage Weil <sage@newdream.net>
2011-03-03 10:09:50 -08:00
Fry, Donald H
f8f79a5dbe iwlagn: report correct temperature for WiFi/BT devices.
The temperature reported by 'cat /sys/class/net/wlan?/device/temperature'
is incorrect for devices with BT capability.  Report the value from the
correct statistics structure.  Tested with 130, 100, 6205 and 5300.

Signed-off-by:  Don Fry <donald.h.fry@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
2011-03-03 10:05:11 -08:00
Konrad Rzeszutek Wilk
1aa0b51a03 xen/irq: Cleanup up the pirq_to_irq for DomU PV PCI passthrough guests as well.
We only did this for PV guests that are xen_initial_domain() but
there is not reason not to do this for other cases. The other
case is only exercised when you pass in a PCI device to a PV guest
_and_ the device in question.

Reviewed-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 12:02:22 -05:00
Thomas Gleixner
676dc3cf5b xen: Use IRQF_FORCE_RESUME
Mark the IRQF_NO_SUSPEND interrupts IRQF_FORCE_RESUME and remove the extra
walk through the interrupt descriptors.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 12:02:21 -05:00
Konrad Rzeszutek Wilk
8aef4857d2 Merge branch 'irq/for-xen' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip into HEAD
* 'irq/for-xen' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  genirq: Add IRQF_FORCE_RESUME
2011-03-03 12:02:02 -05:00
Ian Campbell
f611f2da99 xen/timer: Missing IRQF_NO_SUSPEND in timer code broke suspend.
The patches missed an indirect use of IRQF_NO_SUSPEND pulled in via
IRQF_TIMER. The following patch fixes the issue.

With this fixlet PV guest migration works just fine. I also booted the
entire series as a dom0 kernel and it appeared fine.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 12:00:31 -05:00
Ian Campbell
aa673c1cb3 xen: Fix compile error introduced by "switch to new irq_chip functions"
drivers/xen/events.c: In function 'ack_pirq':
drivers/xen/events.c:568: error: implicit declaration of function 'irq_move_irq'

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 12:00:30 -05:00
Thomas Gleixner
c9e265e030 xen: Switch to new irq_chip functions
Convert Xen to the new irq_chip functions. Brings us closer to enable
CONFIG_GENERIC_HARDIRQS_NO_DEPRECATED

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 12:00:29 -05:00
Thomas Gleixner
149f256f8c xen: Remove stale irq_chip.end
irq_chip.end got obsolete with the removal of __do_IRQ()

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 12:00:28 -05:00
Ian Campbell
7214610475 xen: events: do not free legacy IRQs
c514d00c8057 "xen: events: add xen_allocate_irq_{dynamic, gsi} and
xen_free_irq" correctly avoids reallocating legacy IRQs (which are
managed by the arch core) but erroneously did not prevent them being
freed.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 12:00:27 -05:00
Ian Campbell
89911501f3 xen: events: allocate GSIs and dynamic IRQs from separate IRQ ranges.
There are three cases which we need to care about, PV guest, PV domain
0 and HVM guest.

The PV guest case is simple since it has no access to ACPI or real
APICs and therefore has no GSIs therefore we simply dynamically
allocate all IRQs. The potentially interesting case here is PIRQ type
event channels associated with passed through PCI devices. However
even in this case the guest has no direct interaction with the
physical GSI since that happens in the PCI backend.

The PV domain 0 and HVM guest cases are actually the same. In domain 0
case the kernel sees the host ACPI and GSIs (although it only sees the
APIC indirectly via the hypervisor) and in the HVM guest case it sees
the virtualised ACPI and emulated APICs. In these cases we start
allocating dynamic IRQs at nr_irqs_gsi so that they cannot clash with
any GSI.

Currently xen_allocate_irq_dynamic starts at nr_irqs and works
backwards looking for a free IRQ in order to (try and) avoid clashing
with GSIs used in domain 0 and in HVM guests. This change avoids that
although we retain the behaviour of allowing dynamic IRQs to encroach
on the GSI range if no suitable IRQs are available since a future IRQ
clash is deemed preferable to failure right now.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
2011-03-03 12:00:21 -05:00
Ian Campbell
c9df1ce585 xen: events: add xen_allocate_irq_{dynamic, gsi} and xen_free_irq
This is neater than open-coded calls to irq_alloc_desc_at and
irq_free_desc.

No intended behavioural change.

Note that we previously were not checking the return value of
irq_alloc_desc_at which would be failing for GSI<NR_IRQS_LEGACY
because the core architecture code has already allocated those for
us. Hence the additional check against NR_IRQS_LEGACY in
xen_allocate_irq_gsi.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
2011-03-03 11:56:59 -05:00
Ian Campbell
cbf6aa89fc xen:events: move find_unbound_irq inside CONFIG_PCI_MSI
The only caller is xen_allocate_pirq_msi which is also under this
ifdef so this fixes:
    drivers/xen/events.c:377: warning: 'find_unbound_pirq' defined but not used
when CONFIG_PCI_MSI=n

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
2011-03-03 11:56:58 -05:00
Ian Campbell
3f2a230caf xen: handled remapped IRQs when enabling a pcifront PCI device.
This happens to not be an issue currently because we take pains to try
to ensure that the GSI-IRQ mapping is 1-1 in a PV guest and that
regular event channels do not clash. However a subsequent patch is
going to break this 1-1 mapping.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
2011-03-03 11:56:57 -05:00
Martin Schwidefsky
0c2bd9b24e [S390] tape: deadlock on system work queue
The 34xx and 3590 tape driver uses the system work queue to defer work
from the interrupt function to process context, e.g. a medium sense
after an unsolicited interrupt. The tape commands started by the work
handler need to be asynchronous, otherwise a deadlock on the system
work queue can occur.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2011-03-03 17:56:14 +01:00
Dan Carpenter
b652277b09 [S390] keyboard: integer underflow bug
The "ct" variable should be an unsigned int.  Both struct kbdiacrs
->kb_cnt and struct kbd_data ->accent_table_size are unsigned ints.

Making it signed causes a problem in KBDIACRUC because the user could
set the signed bit and cause a buffer overflow.

Cc: <stable@kernel.org>
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2011-03-03 17:56:14 +01:00
Martin Schwidefsky
0c0db0355b [S390] xpram: remove __initdata attribute from module parameters
The module parameter 'devs' and 'sizes' are marked as __initdata. The
memory for the parameters are freed after module_init completed. This
can lead to kernel crashes in param_free_charp. Remove the __initdata
attribute to fix the problem.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2011-03-03 17:56:14 +01:00
Petr Uzel
fd51469fb6 block: kill loop_mutex
Following steps lead to deadlock in kernel:

dd if=/dev/zero of=img bs=512 count=1000
losetup -f img
mkfs.ext2 /dev/loop0
mount -t ext2 -o loop /dev/loop0 mnt
umount mnt/

Stacktrace:
[<c102ec04>] irq_exit+0x36/0x59
[<c101502c>] smp_apic_timer_interrupt+0x6b/0x75
[<c127f639>] apic_timer_interrupt+0x31/0x38
[<c101df88>] mutex_spin_on_owner+0x54/0x5b
[<fe2250e9>] lo_release+0x12/0x67 [loop]
[<c10c4eae>] __blkdev_put+0x7c/0x10c
[<c10a4da5>] fput+0xd5/0x1aa
[<fe2250cf>] loop_clr_fd+0x1a9/0x1b1 [loop]
[<fe225110>] lo_release+0x39/0x67 [loop]
[<c10c4eae>] __blkdev_put+0x7c/0x10c
[<c10a59d9>] deactivate_locked_super+0x17/0x36
[<c10b6f37>] sys_umount+0x27e/0x2a5
[<c10b6f69>] sys_oldumount+0xb/0xe
[<c1002897>] sysenter_do_call+0x12/0x26
[<ffffffff>] 0xffffffff

Regression since 2a48fc0ab2, which introduced the private
loop_mutex as part of the BKL removal process.

As per [1], the mutex can be safely removed.

[1] http://www.gossamer-threads.com/lists/linux/kernel/1341930

Addresses: https://bugzilla.novell.com/show_bug.cgi?id=669394
Addresses: https://bugzilla.kernel.org/show_bug.cgi?id=29172

Signed-off-by: Petr Uzel <petr.uzel@suse.cz>
Cc: stable@kernel.org
Reviewed-by: Nikanth Karthikesan <knikanth@suse.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-03-03 11:53:25 -05:00
Konrad Rzeszutek Wilk
6eaa412f27 xen: Mark all initial reserved pages for the balloon as INVALID_P2M_ENTRY.
With this patch, we diligently set regions that will be used by the
balloon driver to be INVALID_P2M_ENTRY and under the ownership
of the balloon driver. We are OK using the __set_phys_to_machine
as we do not expect to be allocating any P2M middle or entries pages.
The set_phys_to_machine has the side-effect of potentially allocating
new pages and we do not want that at this stage.

We can do this because xen_build_mfn_list_list will have already
allocated all such pages up to xen_max_p2m_pfn.

We also move the check for auto translated physmap down the
stack so it is present in __set_phys_to_machine.

[v2: Rebased with mmu->p2m code split]
Reviewed-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
2011-03-03 11:52:48 -05:00
Fry, Donald H
fea41cc9b1 iwlagn: Support new 1000 microcode.
iwlagn: Support new 1000 microcode.

New iwlwifi-1000 microcode requires driver support for API version 5.
(There is no version 4)

Signed-off-by: Don Fry <donald.h.fry@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
2011-03-03 08:38:14 -08:00
Tao Ma
2d3a8497f8 blktrace: Remove blk_fill_rwbs_rq.
If we enable trace events to trace block actions, We use
blk_fill_rwbs_rq to analyze the corresponding actions
in request's cmd_flags, but we only choose the minor 2 bits
from it, so most of other flags(e.g, REQ_SYNC) are missing.
For example, with a sync write we get:
write_test-2409  [001]   160.013869: block_rq_insert: 3,64 W 0 () 258135 + =
8 [write_test]

Since now we have integrated the flags of both bio and request,
it is safe to pass rq->cmd_flags directly to blk_fill_rwbs and
blk_fill_rwbs_rq isn't needed any more.

With this patch, after a sync write we get:
write_test-2417  [000]   226.603878: block_rq_insert: 3,64 WS 0 () 258135 +=
 8 [write_test]

Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Acked-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-03-03 10:53:20 -05:00
Helmut Schaa
95a776107a wl12xx: Correctly set up protection if non-GF STAs are present
Set the gf_protection bit when calling ACX_HT_BSS_OPERATION according
to the GF bit passed by mac80211 in ht_operation_mode.

[Added a proper commit message -- Luca]

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:20:34 +02:00
Arik Nemtsov
24225b37bd wl12xx: wakeup chip from ELP during scan
Commands are sometimes sent to FW on scan completion. Make sure the chip
is awake to receive them. Sending commands while the chip is in ELP
can cause SDIO read errors and/or crash the FW.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:20:25 +02:00
Takashi Iwai
38c0764190 ALSA: hda - Don't set to D3 in Cirrus errata init verbs
The errata init verbs for CS42xx codecs contain the verbs to set
the power-state of SPDIF nodes to D3, which seem to break the SPDIF
output on some MacBooks.  Since this is executed during the power-up
initialization, we shouldn't turn them down there.

Reported-by: Arun Raghavan <arun.raghavan@collabora.co.uk>
Cc: <stable@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
2011-03-03 15:16:33 +01:00
Ido Yariv
b16d4b6864 wl12xx: Modify requested number of memory blocks
Tests have shown that the requested number of memory blocks is
sub-optimal. Slightly modify the requested number of memory blocks for
TX.

Signed-off-by: Ido Yariv <ido@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:12:57 +02:00
Ido Yariv
b07d403705 wl12xx: Avoid redundant TX work
TX might be handled in the threaded IRQ handler, in which case, TX work
might be scheduled just to discover it has nothing to do.

Save a few context switches by cancelling redundant TX work in case TX
is about to be handled in the threaded IRQ handler. Also, avoid
scheduling TX work from wl1271_op_tx if not needed.

Signed-off-by: Ido Yariv <ido@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:12:57 +02:00
Ido Yariv
2da69b890f wl12xx: Switch to level trigger interrupts
The interrupt of the wl12xx is a level interrupt in nature, since the
interrupt line is not auto-reset. However, since resetting the interrupt
requires bus transactions, this cannot be done from an interrupt
context. Thus, requesting a level interrupt would require to disable the
irq and re-enable it after the HW is acknowledged. Since we now request
a threaded irq, this can also be done by specifying the IRQF_ONESHOT
flag.

Triggering on an edge can be problematic in some platforms, if the
sampling frequency is not sufficient for detecting very frequent
interrupts. In case an interrupt is missed, the driver will hang as the
interrupt line will stay high until it is acknowledged by the driver,
which will never happen.

Fix this by requesting a level triggered interrupt, with the
IRQF_ONESHOT flag.

Signed-off-by: Ido Yariv <ido@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:10:46 +02:00
Ido Yariv
a620865edf wl12xx: Switch to a threaded interrupt handler
To achieve maximal throughput, it is very important to react to
interrupts as soon as possible. Currently the interrupt handler wakes up
a worker for handling interrupts in process context. A cleaner and more
efficient design would be to request a threaded interrupt handler.  This
handler's priority is very high, and can do blocking operations such as
SDIO/SPI transactions.

Some work can be deferred, mostly calls to mac80211 APIs
(ieee80211_rx_ni and ieee80211_tx_status). By deferring such work to a
different worker, we can keep the irq handler thread more I/O
responsive. In addition, on multi-core systems the two threads can be
scheduled on different cores, which will improve overall performance.

The use of WL1271_FLAG_IRQ_PENDING & WL1271_FLAG_IRQ_RUNNING was
changed. For simplicity, always query the FW for more pending
interrupts. Since there are relatively long bursts of interrupts, the
extra FW status read overhead is negligible. In addition, this enables
registering the IRQ handler with the ONESHOT option.

Signed-off-by: Ido Yariv <ido@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:10:46 +02:00
Ido Yariv
393fb560d3 wl12xx: Change claiming of the SDIO bus
The SDIO bus is claimed and released for each SDIO transaction. In
addition to the few CPU cycles it takes to claim and release the bus, it
may also cause undesired side effects such as the MMC host stopping its
internal clocks.

Since only the wl12xx_sdio driver drives this SDIO card, it is safe to
claim the SDIO host once (on power on), and release it only when turning
the power off.

This patch was inspired by Juuso Oikarinen's (juuso.oikarinen@nokia.com)
patch "wl12xx: Change claiming of the (SDIO) bus".

Signed-off-by: Ido Yariv <ido@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:10:46 +02:00
Ido Yariv
606ea9fa0b wl12xx: Do end-of-transactions transfers only if needed
On newer hardware revisions, there is no need to write the host's
counter at the end of a RX transaction. The same applies to writing the
number of packets at the end of a TX transaction.

It is generally a good idea to avoid unnecessary SDIO/SPI transfers.
Throughput and CPU usage are improved when avoiding these.

Send the host's RX counter and the TX packet count only if needed, based
on the hardware revision.

[Changed WL12XX_QUIRK_END_OF_TRANSACTION to use BIT(0) -- Luca]

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:10:46 +02:00
Ido Yariv
8aad24642a wl12xx: Reorder data handling in irq_work
The FW has a limited amount of memory for holding frames. In case it
runs out of memory reserved for RX frames, it'll have no other choice
but to drop packets received from the AP. Thus, it is important to
handle RX data interrupts as soon as possible, before handling anything
else.

In addition, since there are enough TX descriptors to go around, it is
better to first send TX frames, and only then handle TX completions.

Fix this by changing the order of function calls in wl1271_irq_work.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
2011-03-03 16:10:46 +02:00