On 07.03.2025 10:20, Luca Fancellu wrote:
> Hi Julien,
>
>> On 7 Mar 2025, at 09:09, Julien Grall <jul...@xen.org> wrote:
>>
>> Hi Luca,
>>
>> On 07/03/2025 07:58, Luca Fancellu wrote:
>>> When Xen is built without HAS_PASSTHROUGH, there are some parts
>>> in arm where iommu_* functions are called in the codebase, but
>>> their implementation is under xen/drivers/passthrough that is
>>> not built.
>>> So provide some stub for these functions in order to build Xen
>>> when !HAS_PASSTHROUGH, which is the case for example on systems
>>> with MPU support.
>>> For gnttab_need_iommu_mapping() in the Arm part, modify the macro
>>> to use IS_ENABLED for the HAS_PASSTHROUGH Kconfig.
>>> Fixes: 0388a5979b21 ("xen/arm: mpu: Introduce choice between MMU and MPU")
>>> Signed-off-by: Luca Fancellu <luca.fance...@arm.com>
>>> ---
>>> v4 changes:
>>> - re-introduced stub for iommu_use_hap_pt, Stefano suggested
>>> it is ok to have it in iommu.h.
>>> - Reworded comment in iommu_domain_init from Jan suggestion
>>> v3 Changes:
>>> - removed stub for iommu_use_hap_pt, another solution will be
>>> done for the instance in common arm code.
>>> - Moved a comment close to the macro it was referred to
>>> - add comment to iommu_domain_init() stub
>>> - modified commit message
>>> - Add fixes tag
>>> v2 Changes:
>>> - modify gnttab_need_iommu_mapping to use IS_ENABLED
>>> - removed macro that didn't allow some of the parameter to be
>>> evaluated
>>> - Changed commit message
>>> ---
>>> ---
>>> xen/arch/arm/include/asm/grant_table.h | 5 +--
>>> xen/include/xen/iommu.h | 50 +++++++++++++++++++++++++-
>>> 2 files changed, 52 insertions(+), 3 deletions(-)
>>> diff --git a/xen/arch/arm/include/asm/grant_table.h
>>> b/xen/arch/arm/include/asm/grant_table.h
>>> index d3c518a926b9..c5d87b60c4df 100644
>>> --- a/xen/arch/arm/include/asm/grant_table.h
>>> +++ b/xen/arch/arm/include/asm/grant_table.h
>>> @@ -73,8 +73,9 @@ int replace_grant_host_mapping(uint64_t gpaddr, mfn_t
>>> frame,
>>> #define gnttab_status_gfn(d, t, i) \
>>> page_get_xenheap_gfn(gnttab_status_page(t, i))
>>> -#define gnttab_need_iommu_mapping(d) \
>>> - (is_domain_direct_mapped(d) && is_iommu_enabled(d))
>>> +#define gnttab_need_iommu_mapping(d) \
>>> + (IS_ENABLED(CONFIG_HAS_PASSTHROUGH) && is_domain_direct_mapped(d) && \
>>> + is_iommu_enabled(d))
>>> #endif /* __ASM_GRANT_TABLE_H__ */
>>> /*
>>> diff --git a/xen/include/xen/iommu.h b/xen/include/xen/iommu.h
>>> index 77a514019cc6..5ac038521e23 100644
>>> --- a/xen/include/xen/iommu.h
>>> +++ b/xen/include/xen/iommu.h
>>> @@ -110,6 +110,8 @@ extern int8_t iommu_hwdom_reserved;
>>> extern unsigned int iommu_dev_iotlb_timeout;
>>> +#ifdef CONFIG_HAS_PASSTHROUGH
>>> +
>>> int iommu_setup(void);
>>> int iommu_hardware_setup(void);
>>> @@ -122,6 +124,28 @@ int arch_iommu_domain_init(struct domain *d);
>>> void arch_iommu_check_autotranslated_hwdom(struct domain *d);
>>> void arch_iommu_hwdom_init(struct domain *d);
>>> +#else
>>> +
>>> +static inline int iommu_setup(void)
>>> +{
>>> + return -ENODEV;
>>> +}
>>> +
>>> +static inline int iommu_domain_init(struct domain *d, unsigned int opts)
>>> +{
>>> + /*
>>> + * Return as the real iommu_domain_init() would: Success when
>>> + * !is_iommu_enabled(), following from !iommu_enabled when
>>> !HAS_PASSTHROUGH
>>> + */
>>> + return 0;
>>> +}
>>> +
>>> +static inline void iommu_hwdom_init(struct domain *d) {}
>>> +
>>> +static inline void iommu_domain_destroy(struct domain *d) {}
>>> +
>>> +#endif /* HAS_PASSTHROUGH */
>>> +
>>> /*
>>> * The following flags are passed to map (applicable ones also to unmap)
>>> * operations, while some are passed back by lookup operations.
>>> @@ -209,6 +233,8 @@ struct msi_msg;
>>> #ifdef CONFIG_HAS_DEVICE_TREE
>>> #include <xen/device_tree.h>
>>> +#ifdef CONFIG_HAS_PASSTHROUGH
>>> +
>>> int iommu_assign_dt_device(struct domain *d, struct dt_device_node *dev);
>>> int iommu_deassign_dt_device(struct domain *d, struct dt_device_node *dev);
>>> int iommu_dt_domain_init(struct domain *d);
>>> @@ -238,6 +264,26 @@ int iommu_do_dt_domctl(struct xen_domctl *domctl,
>>> struct domain *d,
>>> */
>>> int iommu_remove_dt_device(struct dt_device_node *np);
>>> +#else
>>> +
>>> +static inline int iommu_assign_dt_device(struct domain *d,
>>> + struct dt_device_node *dev)
>>> +{
>>> + return -EINVAL;
>>> +}
>>> +
>>> +static inline int iommu_add_dt_device(struct dt_device_node *np)
>>> +{
>>> + return 1;
>>
>> I would suggest to add a comment explain what 1 means. IIRC, this means "no
>> iommu" present.
>
> Would it be ok something like in iommu_domain_init:
>
> /*
> * Returns as the real iommu_add_dt_device() would: Error “no iommu" because
> * !iommu_enabled due to the fact that !HAS_PASSTHROUGH
> */
We had been there before, hadn't we? Personally I find the suggested text
hard to follow. How about
/*
* !HAS_PASSTHROUGH => !iommu_enabled (see the non-stub
* iommu_add_dt_device())
*/
If that's too terse, adding a few words should be fine.
Jan
