On Thursday 22 May 2008 10:48:29 am Moore, Robert wrote: > This has already been addressed by two patches to the ACPICA code. See > bugzilla 9487. Oh, yes, I see the ACPICA fix in the code, now that you point it out. Unfortunately, nobody's hooked up the Linux side of things yet, so we're still broken. I'll try to post a patch for that today. > >-----Original Message----- > >From: Bjorn Helgaas [mailto:bjorn.helgaas@xxxxxx] > >Sent: Wednesday, May 21, 2008 4:25 PM > >To: Len Brown; Moore, Robert > >Cc: linux-acpi@xxxxxxxxxxxxxxx > >Subject: ACPI CA resource encoding problem > > > >I've tripped over some devices that work with PNPBIOS but not with > >PNPACPI, even though Linux is configuring the devices identically. > > > >Linux evaluates _CRS to get a template, which it then fills in and > >passes to _SRS to configure the device. The ACPI CA decodes the AML > >bytestream returned from _CRS into a series of acpi_resource structs. > > > >When Linux evaluates _SRS, the ACPI CA re-encodes the acpi_resource > >structs into an AML bytestream. However, the encoded bytestream > >sometimes has a different form than that returned from _CRS. > > > >For example, acpi_rs_set_irq encodes a small IRQ descriptor. If the > >IRQ is edge-sensitive, active high, and exclusive, acpi_rs_set_irq > >optimizes the output AML by omitting the optional flags byte. > > > >I think it would be better for acpi_rs_set_irq to generate an AML > >descriptor that exactly matches the format and size of the descriptor > >we got from _CRS because _SRS may not be smart enough to handle the > >variation. > > > >On a Toshiba Portege 4000, the parport_pc and smsc-ircc2 drivers both > >work with PNPBIOS, and neither works with PNPACPI. I think the reason > >is this acpi_rs_set_irq optimization. If I apply the attached patch, > >which removes that optimization, both drivers work with PNPACPI. > > > >I don't claim that this patch is correct, because it *always* generates > >3-byte IRQ descriptors, while _SRS may be expecting a 2-byte > descriptor. > >But I think it is enough to point out a problem. > > > >Bjorn > > > > > >Index: work10/drivers/acpi/resources/rsirq.c > >=================================================================== > >--- work10.orig/drivers/acpi/resources/rsirq.c 2008-05-21 > >15:14:10.000000000 -0600 > >+++ work10/drivers/acpi/resources/rsirq.c 2008-05-21 > 15:15:11.000000000 - > >0600 > >@@ -152,26 +152,6 @@ struct acpi_rsconvert_info acpi_rs_set_i > > /* Reset length to 3 bytes (descriptor with flags byte) */ > > > > {ACPI_RSC_LENGTH, 0, 0, sizeof(struct aml_resource_irq)}, > >- > >- /* > >- * Check if the flags byte is necessary. Not needed if the flags > are: > >- * ACPI_EDGE_SENSITIVE, ACPI_ACTIVE_HIGH, ACPI_EXCLUSIVE > >- */ > >- {ACPI_RSC_EXIT_NE, ACPI_RSC_COMPARE_VALUE, > >- ACPI_RS_OFFSET(data.irq.triggering), > >- ACPI_EDGE_SENSITIVE}, > >- > >- {ACPI_RSC_EXIT_NE, ACPI_RSC_COMPARE_VALUE, > >- ACPI_RS_OFFSET(data.irq.polarity), > >- ACPI_ACTIVE_HIGH}, > >- > >- {ACPI_RSC_EXIT_NE, ACPI_RSC_COMPARE_VALUE, > >- ACPI_RS_OFFSET(data.irq.sharable), > >- ACPI_EXCLUSIVE}, > >- > >- /* We can optimize to a 2-byte irq_no_flags() descriptor */ > >- > >- {ACPI_RSC_LENGTH, 0, 0, sizeof(struct aml_resource_irq_noflags)} > > }; > > > > > >/********************************************************************** > **** > >***** > -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
