mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Vijayanand Jitta <vijayanand.jitta@oss.qualcomm.com>
To: "Neil Armstrong" <neil.armstrong@linaro.org>,
	"Nipun Gupta" <nipun.gupta@amd.com>,
	"Nikhil Agarwal" <nikhil.agarwal@amd.com>,
	"Joerg Roedel" <joro@8bytes.org>, "Will Deacon" <will@kernel.org>,
	"Robin Murphy" <robin.murphy@arm.com>,
	"Lorenzo Pieralisi" <lpieralisi@kernel.org>,
	"Marc Zyngier" <maz@kernel.org>,
	"Thomas Gleixner" <tglx@kernel.org>,
	"Rob Herring" <robh@kernel.org>,
	"Saravana Kannan" <saravanak@kernel.org>,
	"Richard Zhu" <hongxing.zhu@nxp.com>,
	"Lucas Stach" <l.stach@pengutronix.de>,
	"Krzysztof Wilczyński" <kwilczynski@kernel.org>,
	"Manivannan Sadhasivam" <mani@kernel.org>,
	"Bjorn Helgaas" <bhelgaas@google.com>,
	"Frank Li" <Frank.Li@nxp.com>,
	"Sascha Hauer" <s.hauer@pengutronix.de>,
	"Pengutronix Kernel Team" <kernel@pengutronix.de>,
	"Fabio Estevam" <festevam@gmail.com>,
	"Juergen Gross" <jgross@suse.com>,
	"Stefano Stabellini" <sstabellini@kernel.org>,
	"Oleksandr Tyshchenko" <oleksandr_tyshchenko@epam.com>
Cc: linux-arm-msm@vger.kernel.org, linux-kernel@vger.kernel.org,
	iommu@lists.linux.dev, linux-arm-kernel@lists.infradead.org,
	devicetree@vger.kernel.org, linux-pci@vger.kernel.org,
	imx@lists.linux.dev, xen-devel@lists.xenproject.org,
	Charan Teja Kalla <charan.kalla@oss.qualcomm.com>
Subject: Re: [PATCH v16 3/3] of: Respect #{iommu,msi}-cells in maps
Date: Tue, 28 Jul 2026 09:35:58 +0530	[thread overview]
Message-ID: <0d0896af-a4a5-445d-8db7-fdb9eecd07a2@oss.qualcomm.com> (raw)
In-Reply-To: <3f5c974f-425d-47a5-9fda-e05de1f39d79@linaro.org>



On 7/23/2026 6:47 PM, Neil Armstrong wrote:
> Hi,
> 
> On 6/3/26 09:13, Vijayanand Jitta wrote:
>> From: Robin Murphy <robin.murphy@arm.com>
>>
>> So far our parsing of {iommu,msi}-map properties has always blindly
>> assumed that the output specifiers will always have exactly 1 cell.
>> This typically does happen to be the case, but is not actually enforced
>> (and the PCI msi-map binding even explicitly states support for 0 or 1
>> cells) - as a result we've now ended up with dodgy DTs out in the field
>> which depend on this behaviour to map a 1-cell specifier for a 2-cell
>> provider, despite that being bogus per the bindings themselves.
>>
>> Since there is some potential use in being able to map at least single
>> input IDs to multi-cell output specifiers (and properly support 0-cell
>> outputs as well), add support for properly parsing and using the target
>> nodes' #cells values, albeit with the unfortunate complication of still
>> having to work around expectations of the old behaviour too.
>>
>> Since there are multi-cell output specifiers, the callers of of_map_id()
>> may need to get the exact cell output value for further processing.
>> Update of_map_id() to set args_count in the output to reflect the actual
>> number of output specifier cells.
>>
>> Signed-off-by: Robin Murphy <robin.murphy@arm.com>
>> Signed-off-by: Charan Teja Kalla <charan.kalla@oss.qualcomm.com>
>> Signed-off-by: Vijayanand Jitta <vijayanand.jitta@oss.qualcomm.com>
>> ---
>>   drivers/of/base.c  | 168 +++++++++++++++++++++++++++++++++++++++++------------
>>   include/linux/of.h |   6 +-
>>   2 files changed, 135 insertions(+), 39 deletions(-)
>>
>> diff --git a/drivers/of/base.c b/drivers/of/base.c
>> index d658c2620135..ac7961cbab94 100644
>> --- a/drivers/of/base.c
>> +++ b/drivers/of/base.c
>> @@ -2116,19 +2116,49 @@ int of_find_last_cache_level(unsigned int cpu)
>>       return cache_level;
>>   }
>>   +/*
>> + * Some DTs have an iommu-map targeting a 2-cell IOMMU node while
>> + * specifying only 1 cell. Fortunately they all consist of value '1'
>> + * as the 2nd cell entry with the same target, so check for that pattern.
>> + *
>> + * Example:
>> + *    IOMMU node:
>> + *        #iommu-cells = <2>;
>> + *
>> + *    Device node:
>> + *        iommu-map = <0x0000 &smmu 0x0000 0x1>,
>> + *                <0x0100 &smmu 0x0100 0x1>;
> 
> So the sm8650 PCIe controllers has:
> 
> pcie@1c08000:
>             iommu-map = <0     &apps_smmu 0x1480 0x1>,
>                     <0x100 &apps_smmu 0x1481 0x1>;
> 
> and
> 
> pcie@1c00000:
> 
>             iommu-map = <0     &apps_smmu 0x1400 0x1>,
>                     <0x100 &apps_smmu 0x1401 0x1>;
> 
> and apps_smmu has #iommu-cells = <2>, but gets flagged at wrong:
> 
> [    7.538800] OF: /soc@0/pcie@1c08000: iommu-map has 1-cell entries targeting 2-cell #iommu-cells, treating as 1-cell output
> 
> Returning false in of_check_bad_map() triggers:
> 
> [    7.642680] OF: /soc@0/pcie@1c08000: Unsupported iommu-map - cannot handle 256-ID range with 2-cell output specifier
> 
> I don't understand the issue here, we use 2 cells as expected by
> the iommu-cells, so why is it wrong ? can somebody explain in
> comprehensive words ? I'm super confused, it worked like a charm until now.
> 
> Neil
> 

Hi Neil,

iommu-map = <0     &apps_smmu 0x1480 0x1>,
            <0x100 &apps_smmu 0x1481 0x1>;


Entries here are not 2-cell format, Even though apps_smmu declares #iommu-cells = <2>,
this DT only supplies one output cell (0x1480/0x1481) — the trailing 0x1 is the length field,
not a second output cell. (<id-base phandle out-base length>)

The new code detects exactly this pattern (same target phandle across all entries, length always 1)
and falls back to treating the map as 1-cell output for backward compatibility — hence the pr_warn_once.
It's harmless and expected, your RIDs still resolve to the correct SIDs (0 → 0x1480, 0x100 → 0x1481).

The second message is a different case and shouldn't be coming from this same map — once the 1-cell
fallback triggers on the first entry, it applies to the whole map, so you shouldn't hit both warnings
together on the same node. That error only fires for a genuine 2-cell output specifier combined with
an id_len > 1, e.g.:

iommu-map = <0x0 &apps_smmu 0x1480 0x1 0x100>;

(<id-base, phandle, out0, out1, length=256>) — which isn't supported, since there's no way to
linearly scale a multi-cell output specifier across a range of IDs.

Are you seeing that second error on the same pcie node, or a different one?
If it's the same node, can you share the exact iommu-map entry that triggers it?


Thanks,
Vijay

>> + */
>> +static bool of_check_bad_map(const __be32 *map, int len)
>> +{
>> +    __be32 phandle = map[1];
>> +
>> +    if (len % 4)
>> +        return false;
>> +    for (int i = 0; i < len; i += 4) {
>> +        if (map[i + 1] != phandle || map[i + 3] != cpu_to_be32(1))
>> +            return false;
>> +    }
>> +    return true;
>> +}
>> +
>>   /**
>>    * of_map_id - Translate an ID through a downstream mapping.
>>    * @np: root complex device node.
>>    * @id: device ID to map.
>>    * @map_name: property name of the map to use.
>> + * @cells_name: property name of target specifier cells.
>>    * @map_mask_name: optional property name of the mask to use.
>>    * @filter_np: pointer to an optional filter node, or NULL to allow bypass.
>>    *    If non-NULL, the map property must exist (-ENODEV if absent). If
>>    *    *filter_np is also non-NULL, only entries targeting that node match.
>>    * @arg: pointer to a &struct of_phandle_args for the result. On success,
>> - *    @arg->args[0] will contain the translated ID. If a map entry was
>> - *    matched, @arg->np will be set to the target node with a reference
>> - *    held that the caller must release with of_node_put().
>> + *    @arg->args_count will be set to the number of output specifier cells
>> + *    as defined by @cells_name in the target node, and
>> + *    @arg->args[0..args_count-1] will contain the translated output
>> + *    specifier values. If a map entry was matched, @arg->np will be set
>> + *    to the target node with a reference held that the caller must release
>> + *    with of_node_put().
>>    *
>>    * Given a device ID, look up the appropriate implementation-defined
>>    * platform ID and/or the target device which receives transactions on that
>> @@ -2137,19 +2167,21 @@ int of_find_last_cache_level(unsigned int cpu)
>>    * Return: 0 on success or a standard error code on failure.
>>    */
>>   int of_map_id(const struct device_node *np, u32 id,
>> -           const char *map_name, const char *map_mask_name,
>> +           const char *map_name, const char *cells_name,
>> +           const char *map_mask_name,
>>              struct device_node * const *filter_np, struct of_phandle_args *arg)
>>   {
>>       u32 map_mask, masked_id;
>> -    int map_len;
>> +    int map_bytes, map_len, offset = 0;
>> +    bool bad_map = false;
>>       const __be32 *map = NULL;
>>   -    if (!np || !map_name || !arg)
>> +    if (!np || !map_name || !cells_name || !arg)
>>           return -EINVAL;
>>       /* Ensure bypass/no-match success never returns a stale target node. */
>>       arg->np = NULL;
>>   -    map = of_get_property(np, map_name, &map_len);
>> +    map = of_get_property(np, map_name, &map_bytes);
>>       if (!map) {
>>           if (filter_np)
>>               return -ENODEV;
>> @@ -2159,11 +2191,9 @@ int of_map_id(const struct device_node *np, u32 id,
>>           return 0;
>>       }
>>   -    if (!map_len || map_len % (4 * sizeof(*map))) {
>> -        pr_err("%pOF: Error: Bad %s length: %d\n", np,
>> -            map_name, map_len);
>> -        return -EINVAL;
>> -    }
>> +    if (map_bytes % sizeof(*map))
>> +        goto err_map_len;
>> +    map_len = map_bytes / sizeof(*map);
>>         /* The default is to select all bits. */
>>       map_mask = 0xffffffff;
>> @@ -2176,39 +2206,93 @@ int of_map_id(const struct device_node *np, u32 id,
>>           of_property_read_u32(np, map_mask_name, &map_mask);
>>         masked_id = map_mask & id;
>> -    for ( ; map_len > 0; map_len -= 4 * sizeof(*map), map += 4) {
>> +
>> +    while (offset < map_len) {
>>           struct device_node *phandle_node;
>> -        u32 id_base = be32_to_cpup(map + 0);
>> -        u32 phandle = be32_to_cpup(map + 1);
>> -        u32 out_base = be32_to_cpup(map + 2);
>> -        u32 id_len = be32_to_cpup(map + 3);
>> +        u32 id_base, phandle, id_len, id_off, cells = 0;
>> +        const __be32 *out_base;
>> +
>> +        if (map_len - offset < 2)
>> +            goto err_map_len;
>> +
>> +        id_base = be32_to_cpup(map + offset);
>>             if (id_base & ~map_mask) {
>> -            pr_err("%pOF: Invalid %s translation - %s-mask (0x%x) ignores id-base (0x%x)\n",
>> -                np, map_name, map_name,
>> -                map_mask, id_base);
>> +            pr_err("%pOF: Invalid %s translation - %s (0x%x) ignores id-base (0x%x)\n",
>> +                   np, map_name, map_mask_name, map_mask, id_base);
>>               return -EFAULT;
>>           }
>>   -        if (masked_id < id_base || masked_id >= id_base + id_len)
>> -            continue;
>> -
>> +        phandle = be32_to_cpup(map + offset + 1);
>>           phandle_node = of_find_node_by_phandle(phandle);
>>           if (!phandle_node)
>>               return -ENODEV;
>>   +        /*
>> +         * Assume 1-cell output specifier if the target node lacks the
>> +         * #cells property, for backward compatibility with controllers
>> +         * that predate the property (e.g. arm,gic-v2m-frame).
>> +         */
>> +        if (bad_map || of_property_read_u32(phandle_node, cells_name, &cells))
>> +            cells = 1;
>> +
>> +        if (cells > MAX_PHANDLE_ARGS) {
>> +            pr_err("%pOF: %s cell count %d exceeds maximum\n",
>> +                   phandle_node, cells_name, cells);
>> +            of_node_put(phandle_node);
>> +            return -EINVAL;
>> +        }
>> +
>> +        if (offset == 0 && cells == 2) {
>> +            bad_map = of_check_bad_map(map, map_len);
>> +            if (bad_map) {
>> +                pr_warn_once("%pOF: %s has 1-cell entries targeting 2-cell %s, treating as 1-cell output\n",
>> +                         np, map_name, cells_name);
>> +                cells = 1;
>> +            }
>> +        }
>> +
>> +        if (map_len - offset < 3 + cells) {
>> +            of_node_put(phandle_node);
>> +            goto err_map_len;
>> +        }
>> +
>> +        out_base = map + offset + 2;
>> +        offset += 3 + cells;
>> +
>> +        id_len = be32_to_cpup(map + offset - 1);
>> +        id_off = masked_id - id_base;
>> +        if (masked_id < id_base || id_off >= id_len) {
>> +            of_node_put(phandle_node);
>> +            continue;
>> +        }
>> +        if (id_len > 1 && cells > 1) {
>> +            /*
>> +             * With 1 output cell we reasonably assume its value
>> +             * has a linear relationship to the input; with more,
>> +             * we'd need help from the provider to know what to do.
>> +             */
>> +            pr_err("%pOF: Unsupported %s - cannot handle %d-ID range with %d-cell output specifier\n",
>> +                   np, map_name, id_len, cells);
>> +            of_node_put(phandle_node);
>> +            return -EINVAL;
>> +        }
>> +
>>           if (filter_np && *filter_np && *filter_np != phandle_node) {
>>               of_node_put(phandle_node);
>>               continue;
>>           }
>>             arg->np = phandle_node;
>> -        arg->args[0] = masked_id - id_base + out_base;
>> -        arg->args_count = 1;
>> +        for (int i = 0; i < cells; i++)
>> +            arg->args[i] = id_off + be32_to_cpu(out_base[i]);
>> +        arg->args_count = cells;
>>             pr_debug("%pOF: %s, using mask %08x, id-base: %08x, out-base: %08x, length: %08x, id: %08x -> %08x\n",
>> -            np, map_name, map_mask, id_base, out_base,
>> -            id_len, id, masked_id - id_base + out_base);
>> +            np, map_name, map_mask, id_base,
>> +            cells ? be32_to_cpup(out_base) : 0,
>> +            id_len, id,
>> +            cells ? id_off + be32_to_cpup(out_base) : id_off);
>>           return 0;
>>       }
>>   @@ -2219,6 +2303,10 @@ int of_map_id(const struct device_node *np, u32 id,
>>       arg->args[0] = id;
>>       arg->args_count = 1;
>>       return 0;
>> +
>> +err_map_len:
>> +    pr_err("%pOF: Error: Bad %s length: %d\n", np, map_name, map_bytes);
>> +    return -EINVAL;
>>   }
>>   EXPORT_SYMBOL_GPL(of_map_id);
>>   @@ -2228,18 +2316,21 @@ EXPORT_SYMBOL_GPL(of_map_id);
>>    * @id: Requester ID of the device (e.g. PCI RID/BDF or a platform
>>    *      stream/device ID) used as the lookup key in the iommu-map table.
>>    * @arg: pointer to a &struct of_phandle_args for the result. On success,
>> - *    @arg->args[0] contains the translated ID. If a map entry was matched,
>> - *    @arg->np holds a reference to the target node that the caller must
>> - *    release with of_node_put().
>> + *    @arg->args_count will be set to the number of output specifier cells
>> + *    and @arg->args[0..args_count-1] will contain the translated output
>> + *    specifier values. If a map entry was matched, @arg->np holds a
>> + *    reference to the target node that the caller must release with
>> + *    of_node_put().
>>    *
>> - * Convenience wrapper around of_map_id() using "iommu-map" and "iommu-map-mask".
>> + * Convenience wrapper around of_map_id() using "iommu-map", "#iommu-cells",
>> + * and "iommu-map-mask".
>>    *
>>    * Return: 0 on success or a standard error code on failure.
>>    */
>>   int of_map_iommu_id(const struct device_node *np, u32 id,
>>               struct of_phandle_args *arg)
>>   {
>> -    return of_map_id(np, id, "iommu-map", "iommu-map-mask", NULL, arg);
>> +    return of_map_id(np, id, "iommu-map", "#iommu-cells", "iommu-map-mask", NULL, arg);
>>   }
>>   EXPORT_SYMBOL_GPL(of_map_iommu_id);
>>   @@ -2252,17 +2343,20 @@ EXPORT_SYMBOL_GPL(of_map_iommu_id);
>>    *    If non-NULL, the map property must exist (-ENODEV if absent). If
>>    *    *filter_np is also non-NULL, only entries targeting that node match.
>>    * @arg: pointer to a &struct of_phandle_args for the result. On success,
>> - *    @arg->args[0] contains the translated ID. If a map entry was matched,
>> - *    @arg->np holds a reference to the target node that the caller must
>> - *    release with of_node_put().
>> + *    @arg->args_count will be set to the number of output specifier cells
>> + *    and @arg->args[0..args_count-1] will contain the translated output
>> + *    specifier values. If a map entry was matched, @arg->np holds a
>> + *    reference to the target node that the caller must release with
>> + *    of_node_put().
>>    *
>> - * Convenience wrapper around of_map_id() using "msi-map" and "msi-map-mask".
>> + * Convenience wrapper around of_map_id() using "msi-map", "#msi-cells",
>> + * and "msi-map-mask".
>>    *
>>    * Return: 0 on success or a standard error code on failure.
>>    */
>>   int of_map_msi_id(const struct device_node *np, u32 id,
>>             struct device_node * const *filter_np, struct of_phandle_args *arg)
>>   {
>> -    return of_map_id(np, id, "msi-map", "msi-map-mask", filter_np, arg);
>> +    return of_map_id(np, id, "msi-map", "#msi-cells", "msi-map-mask", filter_np, arg);
>>   }
>>   EXPORT_SYMBOL_GPL(of_map_msi_id);
>> diff --git a/include/linux/of.h b/include/linux/of.h
>> index ea50b45d9ff7..374b249766a2 100644
>> --- a/include/linux/of.h
>> +++ b/include/linux/of.h
>> @@ -465,7 +465,8 @@ const char *of_prop_next_string(const struct property *prop, const char *cur);
>>   bool of_console_check(const struct device_node *dn, char *name, int index);
>>     int of_map_id(const struct device_node *np, u32 id,
>> -           const char *map_name, const char *map_mask_name,
>> +           const char *map_name, const char *cells_name,
>> +           const char *map_mask_name,
>>              struct device_node * const *filter_np,
>>              struct of_phandle_args *arg);
>>   @@ -950,7 +951,8 @@ static inline void of_property_clear_flag(struct property *p, unsigned long flag
>>   }
>>     static inline int of_map_id(const struct device_node *np, u32 id,
>> -                 const char *map_name, const char *map_mask_name,
>> +                 const char *map_name, const char *cells_name,
>> +                 const char *map_mask_name,
>>                    struct device_node * const *filter_np,
>>                    struct of_phandle_args *arg)
>>   {
>>
> 


  reply	other threads:[~2026-07-28  4:07 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-06-03  7:13 [PATCH v16 0/3] of: parsing of multi " Vijayanand Jitta
2026-06-03  7:13 ` [PATCH v16 1/3] of: Add convenience wrappers for of_map_id() Vijayanand Jitta
2026-06-03  7:13 ` [PATCH v16 2/3] of: Factor arguments passed to of_map_id() into a struct Vijayanand Jitta
2026-06-03  7:13 ` [PATCH v16 3/3] of: Respect #{iommu,msi}-cells in maps Vijayanand Jitta
2026-07-23 13:17   ` Neil Armstrong
2026-07-28  4:05     ` Vijayanand Jitta [this message]
2026-07-28 10:04       ` Robin Murphy
2026-07-28 10:39         ` Vijayanand Jitta
2026-07-28 12:57           ` Neil Armstrong
2026-07-28 11:34     ` Dmitry Baryshkov
2026-07-28 12:58       ` Neil Armstrong
2026-06-12 15:35 ` [PATCH v16 0/3] of: parsing of multi " Rob Herring

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=0d0896af-a4a5-445d-8db7-fdb9eecd07a2@oss.qualcomm.com \
    --to=vijayanand.jitta@oss.qualcomm.com \
    --cc=Frank.Li@nxp.com \
    --cc=bhelgaas@google.com \
    --cc=charan.kalla@oss.qualcomm.com \
    --cc=devicetree@vger.kernel.org \
    --cc=festevam@gmail.com \
    --cc=hongxing.zhu@nxp.com \
    --cc=imx@lists.linux.dev \
    --cc=iommu@lists.linux.dev \
    --cc=jgross@suse.com \
    --cc=joro@8bytes.org \
    --cc=kernel@pengutronix.de \
    --cc=kwilczynski@kernel.org \
    --cc=l.stach@pengutronix.de \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=lpieralisi@kernel.org \
    --cc=mani@kernel.org \
    --cc=maz@kernel.org \
    --cc=neil.armstrong@linaro.org \
    --cc=nikhil.agarwal@amd.com \
    --cc=nipun.gupta@amd.com \
    --cc=oleksandr_tyshchenko@epam.com \
    --cc=robh@kernel.org \
    --cc=robin.murphy@arm.com \
    --cc=s.hauer@pengutronix.de \
    --cc=saravanak@kernel.org \
    --cc=sstabellini@kernel.org \
    --cc=tglx@kernel.org \
    --cc=will@kernel.org \
    --cc=xen-devel@lists.xenproject.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®