linux-next: OOPS at bot time

linux-next: OOPS at bot time

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 3532 bytes --]

Hi all,

My Power7 boot test paniced like this: (next-20100722)

%GQLogic Fibre Channel HBA Driver: 8.03.03-k0
qla2xxx 0002:01:00.2: enabling device (0144 -> 0146)
qla2xxx 0002:01:00.2: Found an ISP8001, irq 35, iobase 0xd000080080014000
------------[ cut here ]------------
kernel BUG at drivers/pci/msi.c:205!
Oops: Exception in kernel mode, sig: 5 [#1]
SMP NR_CPUS=128 NUMA pSeries
last sysfs file: /sys/devices/virtual/tty/ptyz8/uevent
Modules linked in: qla2xxx(+)
NIP: c0000000002fcd54 LR: c000000000048d9c CTR: 0000000000000001
REGS: c00000000278aff0 TRAP: 0700   Not tainted  (2.6.35-rc5-autokern1-next-20100721)
MSR: 8000000000029032 <EE,ME,CE,IR,DR>  CR: 28422488  XER: 20000008
TASK = c000000002008000[2226] 'modprobe' THREAD: c000000002788000 CPU: 12
GPR00: 0000000000000001 c00000000278b270 c0000000009a36d0 c0000000009b8900 
GPR04: c00000000278b2e8 ffffffffffffffff 0000000000000000 0000000000020000 
GPR08: 00000000000033e7 c00000000a38b280 0000000000000000 0000000000000000 
GPR12: 0000000088422488 c00000000f331800 00000fff921750a0 0000000000000000 
GPR16: 0000000010033110 00000000100334b8 0000000000000000 0000000000000000 
GPR20: d000080080018000 0000000000022225 c0000000009f7bb4 0000000000010200 
GPR24: 000000002000020d 0000000000000025 c00000000278b2e0 c00000000278b2e8 
GPR28: 0000000000000001 c00000000d0ac5f8 c000000000af8f00 c00000000a38b280 
NIP [c0000000002fcd54] .read_msi_msg_desc+0x24/0x3c
LR [c000000000048d9c] .rtas_setup_msi_irqs+0x1d8/0x254
Call Trace:
[c00000000278b270] [c000000000048d9c] .rtas_setup_msi_irqs+0x1d8/0x254 (unreliable)
[c00000000278b360] [c00000000002a9cc] .arch_setup_msi_irqs+0x34/0x4c
[c00000000278b3e0] [c0000000002fd3fc] .pci_enable_msix+0x49c/0x4ac
[c00000000278b4c0] [d0000000001a5e30] .qla2x00_request_irqs+0x158/0x5b4 [qla2xxx]
[c00000000278b580] [d0000000001cb41c] .qla2x00_probe_one+0xeac/0x63b0 [qla2xxx]
[c00000000278b6f0] [c0000000002f5c4c] .local_pci_probe+0x34/0x48
[c00000000278b760] [c0000000002f6078] .pci_device_probe+0xe8/0x130
[c00000000278b810] [c00000000036e648] .driver_probe_device+0xdc/0x1a4
[c00000000278b8a0] [c00000000036e7a4] .__driver_attach+0x94/0xd8
[c00000000278b930] [c00000000036dabc] .bus_for_each_dev+0x7c/0xe0
[c00000000278b9e0] [c00000000036e410] .driver_attach+0x28/0x40
[c00000000278ba60] [c00000000036d134] .bus_add_driver+0x144/0x310
[c00000000278bb10] [c00000000036ec28] .driver_register+0xd8/0x198
[c00000000278bbb0] [c0000000002f63d0] .__pci_register_driver+0x60/0x10c
[c00000000278bc50] [d0000000001ca520] .qla2x00_module_init+0x150/0x1a0 [qla2xxx]
[c00000000278bce0] [c00000000000947c] .do_one_initcall+0x80/0x1a8
[c00000000278bd90] [c0000000000a4364] .SyS_init_module+0xd8/0x244
[c00000000278be30] [c00000000000852c] syscall_exit+0x0/0x40
Instruction dump:
ebe1fff8 7c0803a6 4e800020 e9230028 81490030 80090034 81690038 7d400378 
7c005b78 7c000034 5400d97e 78000020 <0b000000> 81690038 e8090030 91640008 
---[ end trace f67a78811ed47c60 ]---
%Gudevd-work[1379]: '/sbin/modprobe -b pci:v00001077d00008001sv00001077sd0000017Fbc0Csc04i00' unexpected exit with status 0x0005

That line number is this:

	BUG_ON(!(entry->msg.address_hi | entry->msg.address_lo |
		 entry->msg.data));

in read_msi_msg_desc().  That BUG_ON was added by commit
2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 ("PCI: MSI: Remove unsafe and
unnecessary hardware access") from the pci tree.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 490 bytes --]

Re: linux-next: OOPS at bot time

[Ben Hutchings]

On Fri, 2010-07-23 at 10:22 +1000, Stephen Rothwell wrote:
> Hi all,
> 
> My Power7 boot test paniced like this: (next-20100722)
> 
> %GQLogic Fibre Channel HBA Driver: 8.03.03-k0
> qla2xxx 0002:01:00.2: enabling device (0144 -> 0146)
> qla2xxx 0002:01:00.2: Found an ISP8001, irq 35, iobase 0xd000080080014000
> ------------[ cut here ]------------
> kernel BUG at drivers/pci/msi.c:205!
[...]
> Call Trace:
> [c00000000278b270] [c000000000048d9c] .rtas_setup_msi_irqs+0x1d8/0x254 (unreliable)
> [c00000000278b360] [c00000000002a9cc] .arch_setup_msi_irqs+0x34/0x4c
> [c00000000278b3e0] [c0000000002fd3fc] .pci_enable_msix+0x49c/0x4ac
[...]
> That line number is this:
> 
> 	BUG_ON(!(entry->msg.address_hi | entry->msg.address_lo |
> 		 entry->msg.data));
> 
> in read_msi_msg_desc().  That BUG_ON was added by commit
> 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 ("PCI: MSI: Remove unsafe and
> unnecessary hardware access") from the pci tree.

I wanted to assert that read_msi_msg_desc() is only used to update
MSI/MSI-X descriptors that have already been generated by Linux.  It
looks like you found an exception.

We could make read_msi_msg() fall back to reading from the hardware, but
I think that what the pSeries code is trying to do - save an MSI message
generated by firmware - is different from what the other callers want.
Instead we could add:

void save_msi_msg(unsigned int irq)
{
	struct irq_desc *desc = irq_to_desc(irq);
	struct msi_desc *entry = get_irq_desc_msi(desc);
	struct msi_msg *msg = &entry->msg;

	/* ...followed by the old implementation of read_msi_msg_desc() */
}

Possibly conditional on something like CONFIG_ARCH_NEEDS_SAVE_MSI_MSG.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: linux-next: OOPS at bot time

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 2133 bytes --]

On Fri, 2010-07-23 at 02:19 +0100, Ben Hutchings wrote:
> On Fri, 2010-07-23 at 10:22 +1000, Stephen Rothwell wrote:
> > Hi all,
> > 
> > My Power7 boot test paniced like this: (next-20100722)
> > 
> > %GQLogic Fibre Channel HBA Driver: 8.03.03-k0
> > qla2xxx 0002:01:00.2: enabling device (0144 -> 0146)
> > qla2xxx 0002:01:00.2: Found an ISP8001, irq 35, iobase 0xd000080080014000
> > ------------[ cut here ]------------
> > kernel BUG at drivers/pci/msi.c:205!
> [...]
> > Call Trace:
> > [c00000000278b270] [c000000000048d9c] .rtas_setup_msi_irqs+0x1d8/0x254 (unreliable)
> > [c00000000278b360] [c00000000002a9cc] .arch_setup_msi_irqs+0x34/0x4c
> > [c00000000278b3e0] [c0000000002fd3fc] .pci_enable_msix+0x49c/0x4ac
> [...]
> > That line number is this:
> > 
> > 	BUG_ON(!(entry->msg.address_hi | entry->msg.address_lo |
> > 		 entry->msg.data));
> > 
> > in read_msi_msg_desc().  That BUG_ON was added by commit
> > 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 ("PCI: MSI: Remove unsafe and
> > unnecessary hardware access") from the pci tree.
> 
> I wanted to assert that read_msi_msg_desc() is only used to update
> MSI/MSI-X descriptors that have already been generated by Linux.  It
> looks like you found an exception.
>
> We could make read_msi_msg() fall back to reading from the hardware, but
> I think that what the pSeries code is trying to do - save an MSI message
> generated by firmware - is different from what the other callers want.
> Instead we could add:
> 
> void save_msi_msg(unsigned int irq)
> {
> 	struct irq_desc *desc = irq_to_desc(irq);
> 	struct msi_desc *entry = get_irq_desc_msi(desc);
> 	struct msi_msg *msg = &entry->msg;
> 
> 	/* ...followed by the old implementation of read_msi_msg_desc() */
> }
> 
> Possibly conditional on something like CONFIG_ARCH_NEEDS_SAVE_MSI_MSG.

Maybe.

But then you end up with read_msi_msg(), which doesn't actually read
anything, which I think is confusing. I'd rather read_msi_msg() read the
message, from the device, and we have another routine which returns the
previously saved msg from the msi_desc.

cheers

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[PATCH] PCI: MSI: Restore read_msi_msg_desc(); add get_cached_msi_msg_desc()

[Ben Hutchings]

commit 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 "PCI: MSI: Remove
unsafe and unnecessary hardware access" changed read_msi_msg_desc() to
return the last MSI message written instead of reading it from the
device, since it may be called while the device is in a reduced
power state.

However, the pSeries platform code really does need to read messages
from the device, since they are initially written by firmware.
Therefore:
- Restore the previous behaviour of read_msi_msg_desc()
- Add new functions get_cached_msi_msg{,_desc}() which return the
  last MSI message written
- Use the new functions where appropriate

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
Compile-tested only.

Ben.

 arch/ia64/kernel/msi_ia64.c    |    2 +-
 arch/ia64/sn/kernel/msi_sn.c   |    2 +-
 arch/x86/kernel/apic/io_apic.c |    2 +-
 drivers/pci/msi.c              |   47 +++++++++++++++++++++++++++++++++++----
 include/linux/msi.h            |    2 +
 5 files changed, 47 insertions(+), 8 deletions(-)

diff --git a/arch/ia64/kernel/msi_ia64.c b/arch/ia64/kernel/msi_ia64.c
index 6c89228..4a746ea 100644
--- a/arch/ia64/kernel/msi_ia64.c
+++ b/arch/ia64/kernel/msi_ia64.c
@@ -25,7 +25,7 @@ static int ia64_set_msi_irq_affinity(unsigned int irq,
 	if (irq_prepare_move(irq, cpu))
 		return -1;
 
-	read_msi_msg(irq, &msg);
+	get_cached_msi_msg(irq, &msg);
 
 	addr = msg.address_lo;
 	addr &= MSI_ADDR_DEST_ID_MASK;
diff --git a/arch/ia64/sn/kernel/msi_sn.c b/arch/ia64/sn/kernel/msi_sn.c
index ebfdd6a..0c72dd4 100644
--- a/arch/ia64/sn/kernel/msi_sn.c
+++ b/arch/ia64/sn/kernel/msi_sn.c
@@ -175,7 +175,7 @@ static int sn_set_msi_irq_affinity(unsigned int irq,
 	 * Release XIO resources for the old MSI PCI address
 	 */
 
-	read_msi_msg(irq, &msg);
+	get_cached_msi_msg(irq, &msg);
         sn_pdev = (struct pcidev_info *)sn_irq_info->irq_pciioinfo;
 	pdev = sn_pdev->pdi_linux_pcidev;
 	provider = SN_PCIDEV_BUSPROVIDER(pdev);
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index e41ed24..4dc0084 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -3397,7 +3397,7 @@ static int set_msi_irq_affinity(unsigned int irq, const struct cpumask *mask)
 
 	cfg = desc->chip_data;
 
-	read_msi_msg_desc(desc, &msg);
+	get_cached_msi_msg_desc(desc, &msg);
 
 	msg.data &= ~MSI_DATA_VECTOR_MASK;
 	msg.data |= MSI_DATA_VECTOR(cfg->vector);
diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
index 4c14f31..69b7be3 100644
--- a/drivers/pci/msi.c
+++ b/drivers/pci/msi.c
@@ -197,9 +197,46 @@ void read_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
 {
 	struct msi_desc *entry = get_irq_desc_msi(desc);
 
-	/* We do not touch the hardware (which may not even be
-	 * accessible at the moment) but return the last message
-	 * written.  Assert that this is valid, assuming that
+	BUG_ON(entry->dev->current_state != PCI_D0);
+
+	if (entry->msi_attrib.is_msix) {
+		void __iomem *base = entry->mask_base +
+			entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE;
+
+		msg->address_lo = readl(base + PCI_MSIX_ENTRY_LOWER_ADDR);
+		msg->address_hi = readl(base + PCI_MSIX_ENTRY_UPPER_ADDR);
+		msg->data = readl(base + PCI_MSIX_ENTRY_DATA);
+	} else {
+		struct pci_dev *dev = entry->dev;
+		int pos = entry->msi_attrib.pos;
+		u16 data;
+
+		pci_read_config_dword(dev, msi_lower_address_reg(pos),
+					&msg->address_lo);
+		if (entry->msi_attrib.is_64) {
+			pci_read_config_dword(dev, msi_upper_address_reg(pos),
+						&msg->address_hi);
+			pci_read_config_word(dev, msi_data_reg(pos, 1), &data);
+		} else {
+			msg->address_hi = 0;
+			pci_read_config_word(dev, msi_data_reg(pos, 0), &data);
+		}
+		msg->data = data;
+	}
+}
+
+void read_msi_msg(unsigned int irq, struct msi_msg *msg)
+{
+	struct irq_desc *desc = irq_to_desc(irq);
+
+	read_msi_msg_desc(desc, msg);
+}
+
+void get_cached_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
+{
+	struct msi_desc *entry = get_irq_desc_msi(desc);
+
+	/* Assert that the cache is valid, assuming that
 	 * valid messages are not all-zeroes. */
 	BUG_ON(!(entry->msg.address_hi | entry->msg.address_lo |
 		 entry->msg.data));
@@ -207,11 +244,11 @@ void read_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
 	*msg = entry->msg;
 }
 
-void read_msi_msg(unsigned int irq, struct msi_msg *msg)
+void get_cached_msi_msg(unsigned int irq, struct msi_msg *msg)
 {
 	struct irq_desc *desc = irq_to_desc(irq);
 
-	read_msi_msg_desc(desc, msg);
+	get_cached_msi_msg_desc(desc, msg);
 }
 
 void write_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
diff --git a/include/linux/msi.h b/include/linux/msi.h
index 6991ab5..91b05c1 100644
--- a/include/linux/msi.h
+++ b/include/linux/msi.h
@@ -14,8 +14,10 @@ struct irq_desc;
 extern void mask_msi_irq(unsigned int irq);
 extern void unmask_msi_irq(unsigned int irq);
 extern void read_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg);
+extern void get_cached_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg);
 extern void write_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg);
 extern void read_msi_msg(unsigned int irq, struct msi_msg *msg);
+extern void get_cached_msi_msg(unsigned int irq, struct msi_msg *msg);
 extern void write_msi_msg(unsigned int irq, struct msi_msg *msg);
 
 struct msi_desc {
-- 
1.6.2.5

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH] PCI: MSI: Restore read_msi_msg_desc(); add get_cached_msi_msg_desc()

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 809 bytes --]

On Fri, 2010-07-23 at 14:56 +0100, Ben Hutchings wrote:
> commit 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 "PCI: MSI: Remove
> unsafe and unnecessary hardware access" changed read_msi_msg_desc() to
> return the last MSI message written instead of reading it from the
> device, since it may be called while the device is in a reduced
> power state.
> 
> However, the pSeries platform code really does need to read messages
> from the device, since they are initially written by firmware.
> Therefore:
> - Restore the previous behaviour of read_msi_msg_desc()
> - Add new functions get_cached_msi_msg{,_desc}() which return the
>   last MSI message written
> - Use the new functions where appropriate
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Looks good to me.

cheers


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [PATCH] PCI: MSI: Restore read_msi_msg_desc(); add get_cached_msi_msg_desc()

[Jesse Barnes]

On Fri, 23 Jul 2010 14:56:28 +0100
Ben Hutchings <bhutchings@solarflare.com> wrote:

> commit 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 "PCI: MSI: Remove
> unsafe and unnecessary hardware access" changed read_msi_msg_desc() to
> return the last MSI message written instead of reading it from the
> device, since it may be called while the device is in a reduced
> power state.
> 
> However, the pSeries platform code really does need to read messages
> from the device, since they are initially written by firmware.
> Therefore:
> - Restore the previous behaviour of read_msi_msg_desc()
> - Add new functions get_cached_msi_msg{,_desc}() which return the
>   last MSI message written
> - Use the new functions where appropriate
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
> ---
> Compile-tested only.
> 

Applied to linux-next with Michael's ack, thanks.  I hope it actually
works, I didn't see a tested-by!

-- 
Jesse Barnes, Intel Open Source Technology Center

Re: [PATCH] PCI: MSI: Restore read_msi_msg_desc(); add get_cached_msi_msg_desc()

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 1234 bytes --]

On Fri, 2010-07-30 at 09:42 -0700, Jesse Barnes wrote:
> On Fri, 23 Jul 2010 14:56:28 +0100
> Ben Hutchings <bhutchings@solarflare.com> wrote:
> 
> > commit 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 "PCI: MSI: Remove
> > unsafe and unnecessary hardware access" changed read_msi_msg_desc() to
> > return the last MSI message written instead of reading it from the
> > device, since it may be called while the device is in a reduced
> > power state.
> > 
> > However, the pSeries platform code really does need to read messages
> > from the device, since they are initially written by firmware.
> > Therefore:
> > - Restore the previous behaviour of read_msi_msg_desc()
> > - Add new functions get_cached_msi_msg{,_desc}() which return the
> >   last MSI message written
> > - Use the new functions where appropriate
> > 
> > Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
> > ---
> > Compile-tested only.
> > 
> 
> Applied to linux-next with Michael's ack, thanks.  I hope it actually
> works, I didn't see a tested-by!

I assume Stephen has been using it, otherwise his linux-next boot tests
will all have been failing. Either way he'll test it when it gets into
linux-next proper.

cheers



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [PATCH] PCI: MSI: Restore read_msi_msg_desc(); add get_cached_msi_msg_desc()

[Benjamin Herrenschmidt]

On Fri, 2010-07-23 at 14:56 +0100, Ben Hutchings wrote:
> commit 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 "PCI: MSI: Remove
> unsafe and unnecessary hardware access" changed read_msi_msg_desc() to
> return the last MSI message written instead of reading it from the
> device, since it may be called while the device is in a reduced
> power state.

Looks reasonable... Jesse ?

Cheers,
Ben.

> However, the pSeries platform code really does need to read messages
> from the device, since they are initially written by firmware.
> Therefore:
> - Restore the previous behaviour of read_msi_msg_desc()
> - Add new functions get_cached_msi_msg{,_desc}() which return the
>   last MSI message written
> - Use the new functions where appropriate
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
> ---
> Compile-tested only.
> 
> Ben.
> 
>  arch/ia64/kernel/msi_ia64.c    |    2 +-
>  arch/ia64/sn/kernel/msi_sn.c   |    2 +-
>  arch/x86/kernel/apic/io_apic.c |    2 +-
>  drivers/pci/msi.c              |   47 +++++++++++++++++++++++++++++++++++----
>  include/linux/msi.h            |    2 +
>  5 files changed, 47 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/ia64/kernel/msi_ia64.c b/arch/ia64/kernel/msi_ia64.c
> index 6c89228..4a746ea 100644
> --- a/arch/ia64/kernel/msi_ia64.c
> +++ b/arch/ia64/kernel/msi_ia64.c
> @@ -25,7 +25,7 @@ static int ia64_set_msi_irq_affinity(unsigned int irq,
>  	if (irq_prepare_move(irq, cpu))
>  		return -1;
>  
> -	read_msi_msg(irq, &msg);
> +	get_cached_msi_msg(irq, &msg);
>  
>  	addr = msg.address_lo;
>  	addr &= MSI_ADDR_DEST_ID_MASK;
> diff --git a/arch/ia64/sn/kernel/msi_sn.c b/arch/ia64/sn/kernel/msi_sn.c
> index ebfdd6a..0c72dd4 100644
> --- a/arch/ia64/sn/kernel/msi_sn.c
> +++ b/arch/ia64/sn/kernel/msi_sn.c
> @@ -175,7 +175,7 @@ static int sn_set_msi_irq_affinity(unsigned int irq,
>  	 * Release XIO resources for the old MSI PCI address
>  	 */
>  
> -	read_msi_msg(irq, &msg);
> +	get_cached_msi_msg(irq, &msg);
>          sn_pdev = (struct pcidev_info *)sn_irq_info->irq_pciioinfo;
>  	pdev = sn_pdev->pdi_linux_pcidev;
>  	provider = SN_PCIDEV_BUSPROVIDER(pdev);
> diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
> index e41ed24..4dc0084 100644
> --- a/arch/x86/kernel/apic/io_apic.c
> +++ b/arch/x86/kernel/apic/io_apic.c
> @@ -3397,7 +3397,7 @@ static int set_msi_irq_affinity(unsigned int irq, const struct cpumask *mask)
>  
>  	cfg = desc->chip_data;
>  
> -	read_msi_msg_desc(desc, &msg);
> +	get_cached_msi_msg_desc(desc, &msg);
>  
>  	msg.data &= ~MSI_DATA_VECTOR_MASK;
>  	msg.data |= MSI_DATA_VECTOR(cfg->vector);
> diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
> index 4c14f31..69b7be3 100644
> --- a/drivers/pci/msi.c
> +++ b/drivers/pci/msi.c
> @@ -197,9 +197,46 @@ void read_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
>  {
>  	struct msi_desc *entry = get_irq_desc_msi(desc);
>  
> -	/* We do not touch the hardware (which may not even be
> -	 * accessible at the moment) but return the last message
> -	 * written.  Assert that this is valid, assuming that
> +	BUG_ON(entry->dev->current_state != PCI_D0);
> +
> +	if (entry->msi_attrib.is_msix) {
> +		void __iomem *base = entry->mask_base +
> +			entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE;
> +
> +		msg->address_lo = readl(base + PCI_MSIX_ENTRY_LOWER_ADDR);
> +		msg->address_hi = readl(base + PCI_MSIX_ENTRY_UPPER_ADDR);
> +		msg->data = readl(base + PCI_MSIX_ENTRY_DATA);
> +	} else {
> +		struct pci_dev *dev = entry->dev;
> +		int pos = entry->msi_attrib.pos;
> +		u16 data;
> +
> +		pci_read_config_dword(dev, msi_lower_address_reg(pos),
> +					&msg->address_lo);
> +		if (entry->msi_attrib.is_64) {
> +			pci_read_config_dword(dev, msi_upper_address_reg(pos),
> +						&msg->address_hi);
> +			pci_read_config_word(dev, msi_data_reg(pos, 1), &data);
> +		} else {
> +			msg->address_hi = 0;
> +			pci_read_config_word(dev, msi_data_reg(pos, 0), &data);
> +		}
> +		msg->data = data;
> +	}
> +}
> +
> +void read_msi_msg(unsigned int irq, struct msi_msg *msg)
> +{
> +	struct irq_desc *desc = irq_to_desc(irq);
> +
> +	read_msi_msg_desc(desc, msg);
> +}
> +
> +void get_cached_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
> +{
> +	struct msi_desc *entry = get_irq_desc_msi(desc);
> +
> +	/* Assert that the cache is valid, assuming that
>  	 * valid messages are not all-zeroes. */
>  	BUG_ON(!(entry->msg.address_hi | entry->msg.address_lo |
>  		 entry->msg.data));
> @@ -207,11 +244,11 @@ void read_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
>  	*msg = entry->msg;
>  }
>  
> -void read_msi_msg(unsigned int irq, struct msi_msg *msg)
> +void get_cached_msi_msg(unsigned int irq, struct msi_msg *msg)
>  {
>  	struct irq_desc *desc = irq_to_desc(irq);
>  
> -	read_msi_msg_desc(desc, msg);
> +	get_cached_msi_msg_desc(desc, msg);
>  }
>  
>  void write_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg)
> diff --git a/include/linux/msi.h b/include/linux/msi.h
> index 6991ab5..91b05c1 100644
> --- a/include/linux/msi.h
> +++ b/include/linux/msi.h
> @@ -14,8 +14,10 @@ struct irq_desc;
>  extern void mask_msi_irq(unsigned int irq);
>  extern void unmask_msi_irq(unsigned int irq);
>  extern void read_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg);
> +extern void get_cached_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg);
>  extern void write_msi_msg_desc(struct irq_desc *desc, struct msi_msg *msg);
>  extern void read_msi_msg(unsigned int irq, struct msi_msg *msg);
> +extern void get_cached_msi_msg(unsigned int irq, struct msi_msg *msg);
>  extern void write_msi_msg(unsigned int irq, struct msi_msg *msg);
>  
>  struct msi_desc {
> -- 
> 1.6.2.5
> 

Re: [PATCH] PCI: MSI: Restore read_msi_msg_desc(); add get_cached_msi_msg_desc()

[Ben Hutchings]

On Tue, 2012-11-20 at 18:20 +1100, Benjamin Herrenschmidt wrote:
> On Fri, 2010-07-23 at 14:56 +0100, Ben Hutchings wrote:
> > commit 2ca1af9aa3285c6a5f103ed31ad09f7399fc65d7 "PCI: MSI: Remove
> > unsafe and unnecessary hardware access" changed read_msi_msg_desc() to
> > return the last MSI message written instead of reading it from the
> > device, since it may be called while the device is in a reduced
> > power state.
> 
> Looks reasonable... Jesse ?
[...]

So reasonable that it was committed a couple of years ago!

Where did you dredge this from?

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.