Re: [ 57/72] genirq: Unmask oneshot irqs when thread was not woken

Am 06.03.2012 um 20:45 schrieb Thomas Gleixner:

On Mon, 5 Mar 2012, Stefan Lippers-Hollmann wrote:

quoted

On Monday 05 March 2012, Sven Joachim wrote:

quoted

On 2012-02-28 02:05 +0100, Greg KH wrote:

quoted

3.2-stable review patch.  If anyone has any objections, please let me know.

Alas, the commit below broke the WiFi on my laptop, where
"iwlist wlan0 scan" now just reports: " wlan0     No scan results".

This is not specific to 3.2, the same problem exists in 3.0.23 and
3.3-rc6 while previous releases worked.

I can confirm this regression on an old PIII based notebook with a 
32 bit BCM4306/3 PCMCIA card:

02:00.0 Network controller [0280]: Broadcom Corporation BCM4306 802.11b/g Wireless LAN Controller [14e4:4320] (rev 03)
        Subsystem: Linksys WPC54G v1 / WPC54GS v1 802.11g Wireless-G Notebook Adapter [1737:4320]
        Flags: bus master, fast devsel, latency 64, IRQ 10
        Memory at 20000000 (32-bit, non-prefetchable) [size=8K]
        Capabilities: [40] Power Management version 2
        Kernel driver in use: b43-pci-bridge

I can observe these differences in the dmesg output (3.2.9 + current 
stable queue-3.2):

The kernel configuration is identical in both cases, the only 
difference is "genirq: Unmask oneshot irqs when thread was not woken"
being applied or not.

--- working	<-- 3.2.9 + current stable queue-3.2, with "genirq: Unmask oneshot irqs when thread was not woken" reverted
+++ broken	<-- 3.2.9 + current stable queue-3.2

@@ -125,6 +125,7 @@ Initializing cgroup subsys freezer
 Initializing cgroup subsys net_cls
 Initializing cgroup subsys blkio
 Initializing cgroup subsys perf_event
+CPU serial number disabled.

I can't see why reverting that genirq patch would cause this.

quoted

 mce: CPU supports 5 MCE banks
 SMP alternatives: switching to UP code
 Freeing SMP alternatives: 8k freed

@@ -137,9 +138,16 @@ SMP disabled
 Performance Events: 
 no APIC, boot with the "lapic" boot parameter to force-enable it.
 no hardware sampling interrupt available.
-Broken PMU hardware detected, using software events only.
+p6 PMU driver.
+... version:                0
+... bit width:              32
+... generic registers:      2
+... value mask:             00000000ffffffff
+... max period:             000000007fffffff
+... fixed-purpose events:   0
+... event mask:             0000000000000003

Ditto.

quoted

 Brought up 1 CPUs
 Total of 1 processors activated (1989.75 BogoMIPS).
 devtmpfs: initialized
 PM: Registering ACPI NVS region at eeff000 (4096 bytes)
 print_constraints: dummy: 

@@ -174,12 +182,12 @@ pci 0000:00:00.0: reg 10: [mem 0xec00000
 type 1 class 0x000604
 pci 0000:00:01.0: supports D1
 type 2 class 0x000607
-pci 0000:00:04.0: reg 10: [mem 0x10000000-0x10000fff]
+pci 0000:00:04.0: reg 10: [mem 0x00000000-0x00000fff]

Even more. 

FWIW, I don't see any such differences on my system (with 3.3-rc6).

There is something extremly fishy.

Cheers,
       Sven