From the comments cpufreq_resume() is called with interrupts off so it shouldn't
be calling blocking_notifier_call_chain().

This is the commit that introduced the bug.
http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=e041c683412d5bf44dc2b109053e3b837b71742d

I sent Alan Stern an email with this bugzilla.

Alan Stern says that the bug is caused by cpufreq and there is a fix in the -mm
kernel.

Continue to get the rwsem.c:20 incident with kernel-2.6.17-1.2548.fc6

Aug 13 09:21:45 localhost kernel: Intel machine check architecture supported.
Aug 13 09:21:45 localhost kernel: Intel machine check reporting enabled on CPU#0.
Aug 13 09:21:45 localhost kernel: BUG: sleeping function called from invalid
context at kernel/rwsem.c:20
Aug 13 09:21:45 localhost kernel: in_atomic():0, irqs_disabled():1
Aug 13 09:21:45 localhost kernel:  [<c04051ee>] show_trace_log_lvl+0x58/0x159

Not sure about timing of -mm stuff....

i also see this with 2.6.17-1.2611.fc6

Still there in 2.6.17-1.2647.fc6, but I think the patch for this has gone into
Linus' tree for 2.6.18.

I see this on 2.6.17-1.3001.fc5 when resuming from S3 suspend.

This is Dave Jones' test kernel, based on 2.6.18rc7git2.

Sony Vaio FS640W, Intel 915GM/GMS/910GML.

Intel machine check architecture supported.
Intel machine check reporting enabled on CPU#0.
Back to C!
BUG: sleeping function called from invalid context at kernel/rwsem.c:20
in_atomic():0, irqs_disabled():1
 [<c0403f8e>] show_trace_log_lvl+0x58/0x171
 [<c040456d>] show_trace+0xd/0x10
 [<c040460a>] dump_stack+0x19/0x1b
 [<c0417a28>] __might_sleep+0x8d/0x95
 [<c042ec6c>] down_read+0x15/0x3f
 [<c0427392>] blocking_notifier_call_chain+0x11/0x2d
 [<c0597b8b>] cpufreq_resume+0xf5/0x11d
 [<c05413f9>] __sysdev_resume+0x23/0x57
 [<c0541560>] sysdev_resume+0x19/0x4b
 [<c0545a7c>] device_power_up+0x8/0xf
 [<c0439d17>] suspend_enter+0x3f/0x5e
 [<c0439e93>] enter_state+0x15d/0x190
 [<c0439f4c>] state_store+0x86/0x9c
 [<c049aa9c>] subsys_attr_store+0x20/0x25
 [<c049ab9f>] sysfs_write_file+0xaa/0xcf
 [<c0464cad>] vfs_write+0xab/0x157
 [<c04651e4>] sys_write+0x3b/0x60
 [<c0402dbf>] syscall_call+0x7/0xb
DWARF2 unwinder stuck at syscall_call+0x7/0xb
Leftover inexact backtrace:
 [<c040456d>] show_trace+0xd/0x10
 [<c040460a>] dump_stack+0x19/0x1b
 [<c0417a28>] __might_sleep+0x8d/0x95
 [<c042ec6c>] down_read+0x15/0x3f
 [<c0427392>] blocking_notifier_call_chain+0x11/0x2d
 [<c0597b8b>] cpufreq_resume+0xf5/0x11d
 [<c05413f9>] __sysdev_resume+0x23/0x57
 [<c0541560>] sysdev_resume+0x19/0x4b
 [<c0545a7c>] device_power_up+0x8/0xf
 [<c0439d17>] suspend_enter+0x3f/0x5e
 [<c0439e93>] enter_state+0x15d/0x190
 [<c0439f4c>] state_store+0x86/0x9c
 [<c049aa9c>] subsys_attr_store+0x20/0x25
 [<c049ab9f>] sysfs_write_file+0xaa/0xcf
 [<c0464cad>] vfs_write+0xab/0x157
 [<c04651e4>] sys_write+0x3b/0x60
 [<c0402dbf>] syscall_call+0x7/0xb
PM: Finishing wakeup.
acpi acpi: resuming
agpgart-intel 0000:00:00.0: resuming
pci 0000:00:02.0: resuming
pci 0000:00:02.1: resuming
PM: Writing back config space on device 0000:00:02.1 at offset 4 (was 0, writing
52000000)
HDA Intel 0000:00:1b.0: resuming
PM: Writing back config space on device 0000:00:1b.0 at offset 1 (was 100006,
writing 100002)
ACPI: PCI Interrupt 0000:00:1b.0[A] -> GSI 16 (level, low) -> IRQ 169

I'm seeing a similar Bug with 2.6.18-1 2200 when resuming from suspend
to disk. 2.6.16 hat suspend working, 2.6.17 brike it completely. Now it
seem to work fine expect for trace below.

swsusp: Need to copy 112010 pages
Intel machine check architecture supported.
Intel machine check reporting enabled on CPU#0.
CPU0: Intel P4/Xeon Extended MCE MSRs (12) available
CPU0: Thermal LVT vector (0xf0) already installed
swsusp: Restoring Highmem
BUG: sleeping function called from invalid context at kernel/rwsem.c:20
in_atomic():0, irqs_disabled():1
 [<c0403f10>] dump_trace+0x69/0x1af
 [<c040406e>] show_trace_log_lvl+0x18/0x2c
 [<c04045e9>] show_trace+0xf/0x11
 [<c0404673>] dump_stack+0x15/0x17
 [<c042ef78>] down_read+0x12/0x1f
 [<c0427923>] blocking_notifier_call_chain+0xe/0x29
 [<c05970df>] cpufreq_resume+0x118/0x13f
 [<c053f3d4>] __sysdev_resume+0x20/0x53
 [<c053f515>] sysdev_resume+0x16/0x47
 [<c05435a1>] device_power_up+0x5/0xa
 [<c0437299>] swsusp_suspend+0x74/0x90
 [<c0437720>] pm_suspend_disk+0x43/0xc9
 [<c04368bc>] enter_state+0x52/0x14f
 [<c0436a3e>] state_store+0x85/0x99
 [<c04969ea>] subsys_attr_store+0x1e/0x22
 [<c0496adc>] sysfs_write_file+0xa6/0xcc
 [<c0461292>] vfs_write+0xa8/0x159
 [<c04617d8>] sys_write+0x41/0x67
 [<c0402d9b>] syscall_call+0x7/0xb
DWARF2 unwinder stuck at syscall_call+0x7/0xb

kernel  2.6.18-1.2849.fc6 (i686) has exactly same behavior when resuming from 
suspend through gnome-power-manager 2.16.0. Machine is IBM Thinkpad T42, 
Pentium-M 1.8GHz, ATI RV350 (Mobility Radeon 9600 M10) graphics card and Intel 
wireless network card (supported by ipw2200 driver). After resume screen is 
left blank, although everything else works, including wireless connection. 
dmesg shows:
[quote]
Intel machine check architecture supported.
Intel machine check reporting enabled on CPU#0.
Back to C!
BUG: sleeping function called from invalid context at kernel/rwsem.c:20
in_atomic():0, irqs_disabled():1
 [<c04051db>] dump_trace+0x69/0x1af
 [<c0405339>] show_trace_log_lvl+0x18/0x2c
 [<c04058ed>] show_trace+0xf/0x11
 [<c04059ea>] dump_stack+0x15/0x17
 [<c04394a2>] down_read+0x12/0x20
 [<c0431601>] blocking_notifier_call_chain+0xe/0x29
 [<c05a9d68>] cpufreq_resume+0x118/0x135
 [<c0551998>] __sysdev_resume+0x20/0x53
 [<c0551adb>] sysdev_resume+0x16/0x47
 [<c0555cbf>] device_power_up+0x5/0xa
 [<c0441959>] suspend_enter+0x3b/0x44
 [<c0441a88>] enter_state+0x126/0x176
 [<c0441b5d>] state_store+0x85/0x99
 [<c04a63ae>] subsys_attr_store+0x1e/0x22
 [<c04a64a1>] sysfs_write_file+0xa7/0xce
 [<c046fa71>] vfs_write+0xa8/0x159
 [<c047009e>] sys_write+0x41/0x67
 [<c0404013>] syscall_call+0x7/0xb
DWARF2 unwinder stuck at syscall_call+0x7/0xb
Leftover inexact backtrace:
 =======================
[unquote]

Alex

This bug has been fixed for me.

I have 2.6.18-1.2849.fc6 installed.  Suspend to RAM and suspend to disk wake up
successfully without this error.

(Sony Vaio VGN-FS640/W)

Tried again today with same IBM Thinkpad T42, same kernel 2.6.18-1.2849.fc6 
(i686). No more stack dumps in dmesg, but exactly same behavior - screen left 
blank while everything else works. Another change is that Linux does not try to 
suspend when I am closing the lid (while it did) and logs this fact:

Dec 15 09:29:09 bozy-nb gnome-power-manager: (root) DPMS blanking screen 
because the lid has been closed on ac power
Dec 15 09:29:32 bozy-nb gnome-power-manager: (root) Turning LCD panel back on 
because laptop lid re-opened

must be an ATI card then. When resuming, following logged to dmesg (note, ATI 
VGA card is device 01:00.0):
Dec  6 15:22:09 bozy-nb kernel: agpgart: Found an AGP 2.0 compliant device at 
0000:00:00.0.
Dec  6 15:22:09 bozy-nb kernel: agpgart: Putting AGP V2 device at 0000:00:00.0 
into 1x mode
Dec  6 15:22:09 bozy-nb kernel: agpgart: Putting AGP V2 device at 0000:01:00.0 
into 1x mode
Dec  6 15:22:10 bozy-nb kernel: [drm] Setting GART location based on new memory 
map
Dec  6 15:22:10 bozy-nb kernel: [drm] Loading R300 Microcode
Dec  6 15:22:10 bozy-nb kernel: [drm] writeback test succeeded in 2 usecs

Does it 'see' two video adapters?

Alex

*** Bug 215156 has been marked as a duplicate of this bug. ***

*** Bug 211590 has been marked as a duplicate of this bug. ***

Alex, can you please open a separate bug (if it's not already reported) for 
your issue. Closing this as it has been reported as fixed in the current 
release.

Looks like this bug isn't really fixed. I got it on every resume-from-disk on a
laptop - please reopen this bug. because all duplicates are already closed, too.

Version-Release number of selected component (if applicable):
kernel-2.6.21-1.3228.fc7

grub.conf:
      kernel /boot/vmlinuz-2.6.21-1.3228.fc7 ro root=LABEL=/ vga=0x317
pci=usepirqmask,noacpi noapic irqpoll 5

(PCI and APIC options are required for non-suddenly-jumping-external-mouse support)


agpgart-via 0000:00:00.0: EARLY resume
BUG: sleeping function called from invalid context at kernel/rwsem.c:20
in_atomic():0, irqs_disabled():1
 [<c0439f26>] down_read+0x12/0x28
 [<c04f05ad>] pci_get_subsys+0x71/0xf3
 [<c052259c>] acpi_pci_link_get_current+0x69/0xc9
 [<c051f6e6>] acpi_bus_get_status+0x32/0x8a
 [<c04f0645>] pci_get_device+0x16/0x19
 [<c04eed1b>] quirk_vialatency+0x18/0xcd
 [<c04ee602>] pci_fixup_device+0x6d/0x78
 [<c04efd3d>] pci_device_resume_early+0x14/0x2a
 [<c055665f>] dpm_power_up+0x93/0xc4
 [<c044595b>] swsusp_suspend+0x4d/0x54
 [<c0445ffe>] pm_suspend_disk+0x101/0x21f
 [<c0444e2e>] enter_state+0x4d/0x1c8
 [<c044502e>] state_store+0x85/0x99
 [<c0444fa9>] state_store+0x0/0x99
 [<c04ab9e6>] subsys_attr_store+0x1e/0x22
 [<c04abb0a>] sysfs_write_file+0xbd/0xe7
 [<c04aba4d>] sysfs_write_file+0x0/0xe7
 [<c0475c97>] vfs_write+0xa8/0x154
 [<c04762a6>] sys_write+0x41/0x67
 [<c0404f70>] syscall_call+0x7/0xb
 =======================
Applying VIA southbridge workaround.

Additional info:
$ cat /proc/cpuinfo 
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 6
model           : 8
model name      : mobile AMD Athlon(tm) XP-M 1800+
stepping        : 0
cpu MHz         : 800.056
cache size      : 256 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 1
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 mtrr pge mca cmov pat pse36
mmx fxsr sse syscall mp mmxext 3dnowext 3dnow up ts fid vid
bogomips        : 1600.82
clflush size    : 32

Peter, your issue isn't cpufreq related and is therefore a different issue. 
Please open a new bug for it.