From: Bruno Prémont on
Hi Benjamin,

On Tue, 29 Dec 2009 21:08:11 "Benjamin Li" wrote:
> Could you try running with the attached patch? This debug patch is
> built against the linux-2.6.31.9 kernel. I think the panic is
> occuring right before a reset has occured due to a TX timeout. To
> see if this is happening, this patch will print hardware state
> information when a TX timeout occurs. If you could run with this
> patch and send the logs when the panic occurs, I would really
> appreciate it.
>
> Thanks again.
>
> -Ben

Sorry for replying only this late but I've been too busy with other
things.

Anyhow, I've been doing some more testing yesterday and today and now
am able to reproduce the/a crash pretty easily.

Either running netconsole and doing 'echo t > /proc/sysrq-trigger' via
SSH on otherwise idle server (from local console nothing bad happens),
but then I have no means to communicate with the kernel (I guess it's
deadlocked somewhere in printk code)

The slightly less easy way to trigger it is with a dummy module that
kind of simulates netconsole behavior but with dummy data (see
attached). I have to have some more traffic (TCP?) going on for the bug
to trigger and tell my module multiple times to push data. This way
server is still accessible via VGA or serial console.

Attached are my 'netbomb.c' (which is a modified netconsole.c) and
full kernel log. This time running a 2.6.33-rc8-git3 kernel, having
forward-ported your patch above (e.g. half of it was already present)


I this time I got the following trace:
[ 134.643292] BUG: unable to handle kernel NULL pointer dereference at (null)
[ 134.643304] IP: [<ffffffffa003edc2>] bnx2_poll_work+0x32/0x13d0 [bnx2]
[ 134.643314] PGD 2a972a067 PUD 2aa245067 PMD 0
[ 134.643319] Oops: 0000 [#1] SMP
[ 134.643323] last sysfs file: /sys/devices/pci0000:00/0000:00:1e.0/0000:01:04.6/class
[ 134.643328] CPU 4
[ 134.643334] Pid: 3226, comm: cat Not tainted 2.6.33-rc8-git3-x86_64 #3 /ProLiant DL360 G5
[ 134.643339] RIP: 0010:[<ffffffffa003edc2>] [<ffffffffa003edc2>] bnx2_poll_work+0x32/0x13d0 [bnx2]
[ 134.643347] RSP: 0018:ffff8802a9643b38 EFLAGS: 00010092
[ 134.643351] RAX: 0000000000000000 RBX: ffff8802afab57c0 RCX: 0000000000000010
[ 134.643355] RDX: 0000000000000000 RSI: ffff8802afab57c0 RDI: ffff8802afab4580
[ 134.643359] RBP: ffff8802a9643cd8 R08: ffff8802af051000 R09: 0000000000000007
[ 134.643363] R10: 000000000000000e R11: 0000000000000000 R12: 0000000000000000
[ 134.643367] R13: 0000000000000010 R14: 0000000000000000 R15: ffff8802afab4580
[ 134.643371] FS: 0000000000000000(0000) GS:ffff880028300000(0063) knlGS:00000000f765f6c0
[ 134.643376] CS: 0010 DS: 002b ES: 002b CR0: 000000008005003b
[ 134.643380] CR2: 0000000000000000 CR3: 00000002a9606000 CR4: 00000000000006e0
[ 134.643384] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 134.643388] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 134.643392] Process cat (pid: 3226, threadinfo ffff8802a9642000, task ffff8802aa14bff0)
[ 134.643396] Stack:
[ 134.643398] 0000000000000070 0000000000000002 0000000000000010 ffff8802afab57c0
[ 134.643404] <0> ffff8802afab4580 0000000300000002 0000000000000000 0000000100000002
[ 134.643410] <0> 0000000000000000 0000000200025220 ffffffff81862e80 0000000000000001
[ 134.643418] Call Trace:
[ 134.643427] [<ffffffff8107e3fe>] ? __alloc_pages_nodemask+0xfe/0x660
[ 134.643433] [<ffffffff811b6de6>] ? msi_set_mask_bit+0x26/0xc0
[ 134.643438] [<ffffffff811b6e8b>] ? unmask_msi_irq+0xb/0x10
[ 134.643443] [<ffffffff8106db54>] ? default_enable+0x24/0x40
[ 134.643448] [<ffffffff8106d9b6>] ? check_irq_resend+0x26/0x70
[ 134.643453] [<ffffffff8106cc23>] ? __enable_irq+0x73/0x80
[ 134.643459] [<ffffffffa004019e>] bnx2_poll_msix+0x3e/0xd0 [bnx2]
[ 134.643465] [<ffffffff8135bcd1>] netpoll_poll+0xe1/0x3c0
[ 134.643470] [<ffffffff8135c168>] netpoll_send_skb+0x118/0x210
[ 134.643475] [<ffffffff8135c45b>] netpoll_send_udp+0x1fb/0x210
[ 134.643480] [<ffffffffa00981c5>] write_msg+0x95/0xd0 [netbomb]
[ 134.643485] [<ffffffffa0098255>] netbomb_write+0x55/0xa4 [netbomb]
[ 134.643492] [<ffffffff810f6581>] proc_reg_write+0x71/0xb0
[ 134.643498] [<ffffffff810ab6cb>] vfs_write+0xcb/0x180
[ 134.643503] [<ffffffff810ab870>] sys_write+0x50/0x90
[ 134.643509] [<ffffffff8102a1a4>] sysenter_dispatch+0x7/0x2b
[ 134.643513] Code: 56 41 55 41 54 53 48 81 ec 78 01 00 00 48 89 bd 80 fe ff ff 48 89 b5 78 fe ff ff 89 95 74 fe ff ff 89 8d 70 fe ff ff 48 8b 46 70 <0f> b7 10 31 c0 80 fa ff 0f 94 c0 01 c2 66 39 96 12 02 00 00 0f
[ 134.643551] RIP [<ffffffffa003edc2>] bnx2_poll_work+0x32/0x13d0 [bnx2]
[ 134.643557] RSP <ffff8802a9643b38>
[ 134.643559] CR2: 0000000000000000
[ 134.643563] ---[ end trace 48bdec67d6d7aadb ]---

Running objdump on kernel compile with debugging symbols this matches:
000000000000ad90 <bnx2_poll_work>:
}
}

static int bnx2_poll_work(struct bnx2 *bp, struct bnx2_napi *bnapi,
int work_done, int budget)
{
ad90: 55 push %rbp
ad91: 48 89 e5 mov %rsp,%rbp
ad94: 41 57 push %r15
ad96: 41 56 push %r14
ad98: 41 55 push %r13
ad9a: 41 54 push %r12
ad9c: 53 push %rbx
ad9d: 48 81 ec 78 01 00 00 sub $0x178,%rsp
ada4: 48 89 bd 80 fe ff ff mov %rdi,-0x180(%rbp)
adab: 48 89 b5 78 fe ff ff mov %rsi,-0x188(%rbp)
adb2: 89 95 74 fe ff ff mov %edx,-0x18c(%rbp)
adb8: 89 8d 70 fe ff ff mov %ecx,-0x190(%rbp)
{
u16 cons;

/* Tell compiler that status block fields can change. */
barrier();
cons = *bnapi->hw_tx_cons_ptr;
adbe: 48 8b 46 70 mov 0x70(%rsi),%rax
adc2: 0f b7 10 movzwl (%rax),%edx
barrier();
if (unlikely((cons & MAX_TX_DESC_CNT) == MAX_TX_DESC_CNT))
cons++;
adc5: 31 c0 xor %eax,%eax
adc7: 80 fa ff cmp $0xff,%dl
adca: 0f 94 c0 sete %al
adcd: 01 c2 add %eax,%edx
int work_done, int budget)
{
struct bnx2_tx_ring_info *txr = &bnapi->tx_ring;
struct bnx2_rx_ring_info *rxr = &bnapi->rx_ring;

if (bnx2_get_hw_tx_cons(bnapi) != txr->hw_tx_cons)
adcf: 66 39 96 12 02 00 00 cmp %dx,0x212(%rsi)
add6: 0f 84 4f 03 00 00 je b12b <bnx2_poll_work+0x39b>


So as already determined bnapi->hw_tx_cons_ptr is NULL... but nothing is
happening after that on network side.

Regards,
Bruno
From: Benjamin Li on
Hi Bruno,

No problems. Thanks for following up with this problem, I really
appreciate all your help.

>From your logs it looks like the device came up using MSI, but in the
MSI-X poll routine was being called:

[ 9.836673] bnx2: eth0: using MSI
....

[ 134.643459] [<ffffffffa004019e>] bnx2_poll_msix+0x3e/0xd0 [bnx2]
[ 134.643465] [<ffffffff8135bcd1>] netpoll_poll+0xe1/0x3c0

which is incorrect. If we are in MSI mode, the bnx2_poll() routine
should be used.

I think what is going on here is that during the bnx2x driver
initialization the current bnx2 driver adds all possible NAPI structures
that map to all the hardware vectors (BNX2_MAX_MSIX_VEC=9) to the NAPI
list in the net_device structure regardless if they are used or not
(Seen in drivers/net/bnx2.c:bnx2_init_napi()). This can cause
uninitialized NAPI structures to be placed on the napi_list. Because
this device is in MSI mode, only 1 vector is initialized. Now, the
problem is triggered when net/core/netpoll.c:poll_napi() is called.
This is because this routine will run through the entire napi_list
calling all the poll routines. In your particular case, it is calling
the poll routine on an uninitialized vector causing the kernel panic.

Please try the patch below to see if it solves your problem. Note, this
only have been compile tested and tested against basic traffic runs.
Unfortunately, I could not reproduce the kernel panic with the
instructions below to verify the patch.

Thanks again for all your help in helping us track this down.

-Ben

On Fri, 2010-02-19 at 00:10 -0800, Bruno Pr�mont wrote:
> Hi Benjamin,
>
> On Tue, 29 Dec 2009 21:08:11 "Benjamin Li" wrote:
> > Could you try running with the attached patch? This debug patch is
> > built against the linux-2.6.31.9 kernel. I think the panic is
> > occuring right before a reset has occured due to a TX timeout. To
> > see if this is happening, this patch will print hardware state
> > information when a TX timeout occurs. If you could run with this
> > patch and send the logs when the panic occurs, I would really
> > appreciate it.
> >
> > Thanks again.
> >
> > -Ben
>
> Sorry for replying only this late but I've been too busy with other
> things.
>
> Anyhow, I've been doing some more testing yesterday and today and now
> am able to reproduce the/a crash pretty easily.
>
> Either running netconsole and doing 'echo t > /proc/sysrq-trigger' via
> SSH on otherwise idle server (from local console nothing bad happens),
> but then I have no means to communicate with the kernel (I guess it's
> deadlocked somewhere in printk code)
>
> The slightly less easy way to trigger it is with a dummy module that
> kind of simulates netconsole behavior but with dummy data (see
> attached). I have to have some more traffic (TCP?) going on for the bug
> to trigger and tell my module multiple times to push data. This way
> server is still accessible via VGA or serial console.
>
> Attached are my 'netbomb.c' (which is a modified netconsole.c) and
> full kernel log. This time running a 2.6.33-rc8-git3 kernel, having
> forward-ported your patch above (e.g. half of it was already present)
>
>
> I this time I got the following trace:
> [ 134.643292] BUG: unable to handle kernel NULL pointer dereference at (null)
> [ 134.643304] IP: [<ffffffffa003edc2>] bnx2_poll_work+0x32/0x13d0 [bnx2]
> [ 134.643314] PGD 2a972a067 PUD 2aa245067 PMD 0
> [ 134.643319] Oops: 0000 [#1] SMP
> [ 134.643323] last sysfs file: /sys/devices/pci0000:00/0000:00:1e.0/0000:01:04.6/class
> [ 134.643328] CPU 4
> [ 134.643334] Pid: 3226, comm: cat Not tainted 2.6.33-rc8-git3-x86_64 #3 /ProLiant DL360 G5
> [ 134.643339] RIP: 0010:[<ffffffffa003edc2>] [<ffffffffa003edc2>] bnx2_poll_work+0x32/0x13d0 [bnx2]
> [ 134.643347] RSP: 0018:ffff8802a9643b38 EFLAGS: 00010092
> [ 134.643351] RAX: 0000000000000000 RBX: ffff8802afab57c0 RCX: 0000000000000010
> [ 134.643355] RDX: 0000000000000000 RSI: ffff8802afab57c0 RDI: ffff8802afab4580
> [ 134.643359] RBP: ffff8802a9643cd8 R08: ffff8802af051000 R09: 0000000000000007
> [ 134.643363] R10: 000000000000000e R11: 0000000000000000 R12: 0000000000000000
> [ 134.643367] R13: 0000000000000010 R14: 0000000000000000 R15: ffff8802afab4580
> [ 134.643371] FS: 0000000000000000(0000) GS:ffff880028300000(0063) knlGS:00000000f765f6c0
> [ 134.643376] CS: 0010 DS: 002b ES: 002b CR0: 000000008005003b
> [ 134.643380] CR2: 0000000000000000 CR3: 00000002a9606000 CR4: 00000000000006e0
> [ 134.643384] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> [ 134.643388] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> [ 134.643392] Process cat (pid: 3226, threadinfo ffff8802a9642000, task ffff8802aa14bff0)
> [ 134.643396] Stack:
> [ 134.643398] 0000000000000070 0000000000000002 0000000000000010 ffff8802afab57c0
> [ 134.643404] <0> ffff8802afab4580 0000000300000002 0000000000000000 0000000100000002
> [ 134.643410] <0> 0000000000000000 0000000200025220 ffffffff81862e80 0000000000000001
> [ 134.643418] Call Trace:
> [ 134.643427] [<ffffffff8107e3fe>] ? __alloc_pages_nodemask+0xfe/0x660
> [ 134.643433] [<ffffffff811b6de6>] ? msi_set_mask_bit+0x26/0xc0
> [ 134.643438] [<ffffffff811b6e8b>] ? unmask_msi_irq+0xb/0x10
> [ 134.643443] [<ffffffff8106db54>] ? default_enable+0x24/0x40
> [ 134.643448] [<ffffffff8106d9b6>] ? check_irq_resend+0x26/0x70
> [ 134.643453] [<ffffffff8106cc23>] ? __enable_irq+0x73/0x80
> [ 134.643459] [<ffffffffa004019e>] bnx2_poll_msix+0x3e/0xd0 [bnx2]
> [ 134.643465] [<ffffffff8135bcd1>] netpoll_poll+0xe1/0x3c0
> [ 134.643470] [<ffffffff8135c168>] netpoll_send_skb+0x118/0x210
> [ 134.643475] [<ffffffff8135c45b>] netpoll_send_udp+0x1fb/0x210
> [ 134.643480] [<ffffffffa00981c5>] write_msg+0x95/0xd0 [netbomb]
> [ 134.643485] [<ffffffffa0098255>] netbomb_write+0x55/0xa4 [netbomb]
> [ 134.643492] [<ffffffff810f6581>] proc_reg_write+0x71/0xb0
> [ 134.643498] [<ffffffff810ab6cb>] vfs_write+0xcb/0x180
> [ 134.643503] [<ffffffff810ab870>] sys_write+0x50/0x90
> [ 134.643509] [<ffffffff8102a1a4>] sysenter_dispatch+0x7/0x2b
> [ 134.643513] Code: 56 41 55 41 54 53 48 81 ec 78 01 00 00 48 89 bd 80 fe ff ff 48 89 b5 78 fe ff ff 89 95 74 fe ff ff 89 8d 70 fe ff ff 48 8b 46 70 <0f> b7 10 31 c0 80 fa ff 0f 94 c0 01 c2 66 39 96 12 02 00 00 0f
> [ 134.643551] RIP [<ffffffffa003edc2>] bnx2_poll_work+0x32/0x13d0 [bnx2]
> [ 134.643557] RSP <ffff8802a9643b38>
> [ 134.643559] CR2: 0000000000000000
> [ 134.643563] ---[ end trace 48bdec67d6d7aadb ]---
>
> Running objdump on kernel compile with debugging symbols this matches:
> 000000000000ad90 <bnx2_poll_work>:
> }
> }
>
> static int bnx2_poll_work(struct bnx2 *bp, struct bnx2_napi *bnapi,
> int work_done, int budget)
> {
> ad90: 55 push %rbp
> ad91: 48 89 e5 mov %rsp,%rbp
> ad94: 41 57 push %r15
> ad96: 41 56 push %r14
> ad98: 41 55 push %r13
> ad9a: 41 54 push %r12
> ad9c: 53 push %rbx
> ad9d: 48 81 ec 78 01 00 00 sub $0x178,%rsp
> ada4: 48 89 bd 80 fe ff ff mov %rdi,-0x180(%rbp)
> adab: 48 89 b5 78 fe ff ff mov %rsi,-0x188(%rbp)
> adb2: 89 95 74 fe ff ff mov %edx,-0x18c(%rbp)
> adb8: 89 8d 70 fe ff ff mov %ecx,-0x190(%rbp)
> {
> u16 cons;
>
> /* Tell compiler that status block fields can change. */
> barrier();
> cons = *bnapi->hw_tx_cons_ptr;
> adbe: 48 8b 46 70 mov 0x70(%rsi),%rax
> adc2: 0f b7 10 movzwl (%rax),%edx
> barrier();
> if (unlikely((cons & MAX_TX_DESC_CNT) == MAX_TX_DESC_CNT))
> cons++;
> adc5: 31 c0 xor %eax,%eax
> adc7: 80 fa ff cmp $0xff,%dl
> adca: 0f 94 c0 sete %al
> adcd: 01 c2 add %eax,%edx
> int work_done, int budget)
> {
> struct bnx2_tx_ring_info *txr = &bnapi->tx_ring;
> struct bnx2_rx_ring_info *rxr = &bnapi->rx_ring;
>
> if (bnx2_get_hw_tx_cons(bnapi) != txr->hw_tx_cons)
> adcf: 66 39 96 12 02 00 00 cmp %dx,0x212(%rsi)
> add6: 0f 84 4f 03 00 00 je b12b <bnx2_poll_work+0x39b>
>
>
> So as already determined bnapi->hw_tx_cons_ptr is NULL... but nothing is
> happening after that on network side.
>
> Regards,
> Bruno
From: Brian Haley on
Hi Ben,

Benjamin Li wrote:
> Hi Bruno,
>
> No problems. Thanks for following up with this problem, I really
> appreciate all your help.
>
>>From your logs it looks like the device came up using MSI, but in the
> MSI-X poll routine was being called:
>
> [ 9.836673] bnx2: eth0: using MSI
> ...
>
> [ 134.643459] [<ffffffffa004019e>] bnx2_poll_msix+0x3e/0xd0 [bnx2]
> [ 134.643465] [<ffffffff8135bcd1>] netpoll_poll+0xe1/0x3c0
>
> which is incorrect. If we are in MSI mode, the bnx2_poll() routine
> should be used.
>
> I think what is going on here is that during the bnx2x driver
> initialization the current bnx2 driver adds all possible NAPI structures
> that map to all the hardware vectors (BNX2_MAX_MSIX_VEC=9) to the NAPI
> list in the net_device structure regardless if they are used or not
> (Seen in drivers/net/bnx2.c:bnx2_init_napi()). This can cause
> uninitialized NAPI structures to be placed on the napi_list. Because
> this device is in MSI mode, only 1 vector is initialized. Now, the
> problem is triggered when net/core/netpoll.c:poll_napi() is called.
> This is because this routine will run through the entire napi_list
> calling all the poll routines. In your particular case, it is calling
> the poll routine on an uninitialized vector causing the kernel panic.
....
> @@ -8201,7 +8204,7 @@ bnx2_init_napi(struct bnx2 *bp)
> {
> int i;
>
> - for (i = 0; i < BNX2_MAX_MSIX_VEC; i++) {
> + for (i = 0; i < bp->irq_nvecs; i++) {
> struct bnx2_napi *bnapi = &bp->bnx2_napi[i];
> int (*poll)(struct napi_struct *, int);

Would this same change need to be made in other places, like bnx2_init_chip()
or bnx2_clear_ring_states() ?

-Brian
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Benjamin Li on
Hi Brian,

On Fri, 2010-02-19 at 13:03 -0800, Brian Haley wrote:
> Hi Ben,
>
> Benjamin Li wrote:
> > Hi Bruno,
> >
> > @@ -8201,7 +8204,7 @@ bnx2_init_napi(struct bnx2 *bp)
> > {
> > int i;
> >
> > - for (i = 0; i < BNX2_MAX_MSIX_VEC; i++) {
> > + for (i = 0; i < bp->irq_nvecs; i++) {
> > struct bnx2_napi *bnapi = &bp->bnx2_napi[i];
> > int (*poll)(struct napi_struct *, int);
>
> Would this same change need to be made in other places, like bnx2_init_chip()
> or bnx2_clear_ring_states() ?

The other locations in the bnx2.c driver are bnx2_init_chip(),
bnx2_clear_ring_states(), bnx2_alloc_mem(). With the current
implementation, the bnx2_napi structures are initialize but never used
which should be ok. But, we can clean this up to save some cycles.

The following are the areas in the code which iterate through all the
vectors.

bnx2_init_chip() - zero the last_status_idx field in the bnx2_napi
structure
bnx2_clear_ring_states() - zero the rings producer/consumer indexes
bnx2_alloc_mem() - initialize the consumer pointers

Thanks again.

-Ben

>
> -Brian
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo(a)vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Bruno Prémont on
Hi Benjamin,

On Fri, 19 February 2010 "Benjamin Li" <benli(a)broadcom.com> wrote:
> >From your logs it looks like the device came up using MSI, but in the
> MSI-X poll routine was being called:
>
> [ 9.836673] bnx2: eth0: using MSI
> ...
>
> [ 134.643459] [<ffffffffa004019e>] bnx2_poll_msix+0x3e/0xd0 [bnx2]
> [ 134.643465] [<ffffffff8135bcd1>] netpoll_poll+0xe1/0x3c0
>
> which is incorrect. If we are in MSI mode, the bnx2_poll() routine
> should be used.
>
> I think what is going on here is that during the bnx2x driver
> initialization the current bnx2 driver adds all possible NAPI
> structures that map to all the hardware vectors (BNX2_MAX_MSIX_VEC=9)
> to the NAPI list in the net_device structure regardless if they are
> used or not (Seen in drivers/net/bnx2.c:bnx2_init_napi()). This can
> cause uninitialized NAPI structures to be placed on the napi_list.
> Because this device is in MSI mode, only 1 vector is initialized.
> Now, the problem is triggered when net/core/netpoll.c:poll_napi() is
> called. This is because this routine will run through the entire
> napi_list calling all the poll routines. In your particular case, it
> is calling the poll routine on an uninitialized vector causing the
> kernel panic.
>
> Please try the patch below to see if it solves your problem. Note,
> this only have been compile tested and tested against basic traffic
> runs. Unfortunately, I could not reproduce the kernel panic with the
> instructions below to verify the patch.
>
> Thanks again for all your help in helping us track this down.

I applied the patch today and tried to reproduce with my showcases.

Seems that it's harder to trigger now but I still end up being able to
crash the box. Don't know if it's the same cause or not (could also
be the tcp-retransmit ghost)...

This time I had to run a few paralell scp's (8Mb/s each) to the box and
'echo t > /proc/sysrq-trigger' multiple times via ssh session for it to
happen. It didn't trigger with by netbomb though I will try some more
and see)

I don't know if it's the same reason or not (hopefully something
reached disk as serial console is dead and pings are not
answered anymore.
It's probably some printk/bug/warn that triggers in network stack and
deadlocks with netconsole.

Regards,
Bruno
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/