Re: [PATCH V7 1/8] ACPI: I/O Remapping Table (IORT) initial support

From: Lorenzo Pieralisi
Date: Tue Jun 21 2016 - 07:23:15 EST


On Mon, Jun 20, 2016 at 01:02:14PM +0200, Tomasz Nowicki wrote:
> IORT shows representation of IO topology for ARM based systems.
> It describes how various components are connected together on
> parent-child basis e.g. PCI RC -> SMMU -> ITS. Also see IORT spec.

Add the spec link if you want a reference to it otherwise "Also see IORT spec"
is pretty much useless.

> Initial support allows to detect IORT table presence and save its
> root pointer obtained through acpi_get_table(). The pointer validity
> depends on acpi_gbl_permanent_mmap because if acpi_gbl_permanent_mmap
> is not set while using IORT nodes we would dereference unmapped pointers.

We still need Rafael's feeback on this.

> For the aforementioned reason call iort_table_detect() from acpi_init()
> which guarantees acpi_gbl_permanent_mmap to be set at that point.

Ditto.

> Add generic helpers which are helpful for scanning and retrieving
> information from IORT table content. List of the most important helpers:
> - iort_find_dev_node() finds IORT node for a given device
> - iort_node_map_rid() maps device RID and returns IORT node which provides
> final translation
>
> Signed-off-by: Tomasz Nowicki <tn@xxxxxxxxxxxx>
> ---
> drivers/acpi/Kconfig | 3 +
> drivers/acpi/Makefile | 1 +
> drivers/acpi/bus.c | 2 +
> drivers/acpi/iort.c | 220 ++++++++++++++++++++++++++++++++++++++++++++++++++
> include/linux/iort.h | 30 +++++++
> 5 files changed, 256 insertions(+)
> create mode 100644 drivers/acpi/iort.c
> create mode 100644 include/linux/iort.h
>
> diff --git a/drivers/acpi/Kconfig b/drivers/acpi/Kconfig
> index b7e2e77..848471f 100644
> --- a/drivers/acpi/Kconfig
> +++ b/drivers/acpi/Kconfig
> @@ -57,6 +57,9 @@ config ACPI_SYSTEM_POWER_STATES_SUPPORT
> config ACPI_CCA_REQUIRED
> bool
>
> +config IORT_TABLE
> + bool
> +
> config ACPI_DEBUGGER
> bool "AML debugger interface"
> select ACPI_DEBUG
> diff --git a/drivers/acpi/Makefile b/drivers/acpi/Makefile
> index 251ce85..c7c9b29 100644
> --- a/drivers/acpi/Makefile
> +++ b/drivers/acpi/Makefile
> @@ -82,6 +82,7 @@ obj-$(CONFIG_ACPI_CUSTOM_METHOD)+= custom_method.o
> obj-$(CONFIG_ACPI_BGRT) += bgrt.o
> obj-$(CONFIG_ACPI_CPPC_LIB) += cppc_acpi.o
> obj-$(CONFIG_ACPI_DEBUGGER_USER) += acpi_dbg.o
> +obj-$(CONFIG_IORT_TABLE) += iort.o
>
> # processor has its own "processor." module_param namespace
> processor-y := processor_driver.o
> diff --git a/drivers/acpi/bus.c b/drivers/acpi/bus.c
> index 31e8da6..176c17d 100644
> --- a/drivers/acpi/bus.c
> +++ b/drivers/acpi/bus.c
> @@ -33,6 +33,7 @@
> #ifdef CONFIG_X86
> #include <asm/mpspec.h>
> #endif
> +#include <linux/iort.h>
> #include <linux/pci.h>
> #include <acpi/apei.h>
> #include <linux/dmi.h>
> @@ -1118,6 +1119,7 @@ static int __init acpi_init(void)
> }
>
> pci_mmcfg_late_init();
> + iort_table_detect();

IORT is arch specific (and empty on any system other than ARM64), this call
here is fine by me because it provides us the ordering we need (ie
initialize the IORT table pointers before we start running scan handlers),
we need Rafael's feeback on this though.

[...]

> +static int
> +iort_id_map(struct acpi_iort_id_mapping *map, u8 type, u32 rid_in, u32 *rid_out)
> +{
> + if (!rid_out)
> + return -EINVAL;

Nit: It is an internal function honestly I do not see the need for
this check here, it is close to paranoia.

> + /* Single mapping does not care for input id */
> + if (map->flags & ACPI_IORT_ID_SINGLE_MAPPING) {
> + if (type == ACPI_IORT_NODE_NAMED_COMPONENT ||
> + type == ACPI_IORT_NODE_PCI_ROOT_COMPLEX) {
> + *rid_out = map->output_base;
> + return 0;
> + }
> +
> + pr_warn(FW_BUG "[map %p] SINGLE MAPPING flag not allowed for node type %d, skipping ID map\n",
> + map, type);
> + return -ENXIO;
> + }
> +
> + if (rid_in < map->input_base ||
> + (rid_in > map->input_base + map->id_count))
> + return -ENXIO;
> +
> + *rid_out = map->output_base + (rid_in - map->input_base);
> + return 0;
> +}
> +
> +static struct acpi_iort_node *
> +iort_node_map_rid(struct acpi_iort_node *node, u32 rid_in,
> + u32 *rid_out, u8 type)
> +{
> + u32 rid = rid_in;
> + int found = 0;
> +
> + /* Climb up ID mapping tree to find specified node type */

s/Climb/Parse the/

> + while (node && node->type != type) {
> + struct acpi_iort_id_mapping *map;
> + int i;
> +
> + if (!node->mapping_offset || !node->mapping_count) {
> + node = NULL;
> + break;
> + }
> +
> + map = ACPI_ADD_PTR(struct acpi_iort_id_mapping, node,
> + node->mapping_offset);
> +
> + /* Firmware bug! */
> + if (!map->output_reference) {
> + pr_err(FW_BUG "[node %p type %d] ID map has NULL parent reference\n",
> + node, node->type);
> + node = NULL;
> + break;
> + }
> +
> + /* Do the RID translation */
> + for (i = 0; i < node->mapping_count; i++, map++) {
> + if(!iort_id_map(map, node->type, rid, &rid)) {
> + found = 1;

You do not reset found to 0, for next iteration. I still think this function
is too complicated.

> + break;
> + }
> + }
> +
> + if (!found) {
> + node = NULL;
> + break;
> + }

You could even remove "found" altogether:

if (i == node->mapping_count)
goto fail_map;

fail_map:
/* Map input RID to output RID unchanged on mapping failure*/
*rid_out = rid_in;
return NULL;

At least it is clear what you return when.

> +
> + node = ACPI_ADD_PTR(struct acpi_iort_node, iort_table,
> + map->output_reference);
> + }
> +
> + /* Take input RID as output RID in case of any error */
> + if (rid_out)
> + *rid_out = node ? rid : rid_in;
> + return node;
> +}
> +
> +static struct acpi_iort_node *
> +iort_find_dev_node(struct device *dev)
> +{
> + struct pci_bus *pbus;
> +
> + if (!dev_is_pci(dev))
> + return iort_scan_node(ACPI_IORT_NODE_NAMED_COMPONENT,
> + iort_match_node_callback, dev);
> +
> + /* Find a PCI root bus */
> + pbus = to_pci_dev(dev)->bus;
> + while (!pci_is_root_bus(pbus))
> + pbus = pbus->parent;
> +
> + return iort_scan_node(ACPI_IORT_NODE_PCI_ROOT_COMPLEX,
> + iort_match_node_callback, &pbus->dev);
> +}
> +
> +void __init iort_table_detect(void)
> +{
> + acpi_status status;
> +
> + if (acpi_disabled)
> + return;

Nit: if you are here acpi_disabled == false, so this check is
pointless.

> + status = acpi_get_table(ACPI_SIG_IORT, 0, &iort_table);
> + if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
> + const char *msg = acpi_format_exception(status);
> + pr_err("Failed to get table, %s\n", msg);
> + }

We really need to know if relying on iort_table to stay mapped
is a valid approach, it seems to me that the answer is yes
given that that's how eg DMAR is handled on x86 but we need
an acknowledgment from Rafael on this.

Other than that and my minor comments above (inclusive of Hanjun's
bug fix for iort_match_node_callback(), I did not reiterate that comment
but we have to fix it) patch seems fine.

Lorenzo

> +}
> diff --git a/include/linux/iort.h b/include/linux/iort.h
> new file mode 100644
> index 0000000..cde6809
> --- /dev/null
> +++ b/include/linux/iort.h
> @@ -0,0 +1,30 @@
> +/*
> + * Copyright (C) 2016, Semihalf
> + * Author: Tomasz Nowicki <tn@xxxxxxxxxxxx>
> + *
> + * This program is free software; you can redistribute it and/or modify it
> + * under the terms and conditions of the GNU General Public License,
> + * version 2, as published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope it will be useful, but WITHOUT
> + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
> + * more details.
> + *
> + * You should have received a copy of the GNU General Public License along with
> + * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
> + * Place - Suite 330, Boston, MA 02111-1307 USA.
> + */
> +
> +#ifndef __IORT_H__
> +#define __IORT_H__
> +
> +#include <linux/acpi.h>
> +
> +#ifdef CONFIG_IORT_TABLE
> +void iort_table_detect(void);
> +#else
> +static inline void iort_table_detect(void) { }
> +#endif
> +
> +#endif /* __IORT_H__ */
> --
> 1.9.1
>