All of lore.kernel.org
 help / color / mirror / Atom feed
From: Robin Murphy <robin.murphy@arm.com>
To: Jason Gunthorpe <jgg@nvidia.com>,
	iommu@lists.linux.dev, Joerg Roedel <joro@8bytes.org>,
	llvm@lists.linux.dev, Nathan Chancellor <nathan@kernel.org>,
	Nick Desaulniers <ndesaulniers@google.com>,
	Miguel Ojeda <ojeda@kernel.org>, Tom Rix <trix@redhat.com>,
	Will Deacon <will@kernel.org>
Cc: Lu Baolu <baolu.lu@linux.intel.com>,
	Kevin Tian <kevin.tian@intel.com>,
	Nicolin Chen <nicolinc@nvidia.com>
Subject: Re: [PATCH v3 08/17] iommu: Make iommu_group_do_dma_first_attach() work with owned groups
Date: Thu, 6 Apr 2023 17:37:57 +0100	[thread overview]
Message-ID: <db92f18c-e2ed-78d0-db7e-83d8ee8a4dce@arm.com> (raw)
In-Reply-To: <8-v3-e89a9bb522f5+8c87-iommu_err_unwind_jgg@nvidia.com>

On 06/04/2023 12:38 am, Jason Gunthorpe wrote:
> If the group is already owned then defered attach should not be done and
> the device should be directly connected to the correct domain.
> 
> Owned means that some driver is already bound to devices in the group and
> is operating the group possibly without DMA API support. In this case
> there would be no way to correct the mismatched domain.

Looking more closely at the current code, it seems we just skip any 
attach at all in this case. Thus although there can be technically be a 
mismatch (physically, the device will be either partly or fully assigned 
to the current domain if it is grouped by ID aliasing, or in limbo if it 
is grouped for ACS reasons), it *should* be corrected whenever the 
current owner next detaches its domain. This shouldn't actually be a 
problem in the usual case, since the new device generally won't do 
anything if the existing ownership prevents its driver from binding.

I don't have a particularly strong opinion on changing this behaviour, 
but I guess what's nagging me is that we could do a better job of 
calling out that it *is* a behaviour change, and its implications. 
Namely that they are slightly better if the device binds to the same 
owner, in that it could rely use the current domain straight off without 
the risk of being in ACS limbo, but also slightly worse in that if the 
attach fails for any reason, it will now trash aliased RIDs of devices 
that *were* working OK. It also means that the device could bind to any 
*other* driver with driver_managed_dma and have a much better guarantee 
of having access to the other owner's domain, with all the pitfalls 
therein, but that's only a strengthening of a pre-existing problem :)

Thanks,
Robin.

> Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
> Reviewed-by: Kevin Tian <kevin.tian@intel.com>
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> ---
>   drivers/iommu/iommu.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
> index c68cf551d05dc4..6fe8bc78db2016 100644
> --- a/drivers/iommu/iommu.c
> +++ b/drivers/iommu/iommu.c
> @@ -411,8 +411,9 @@ static bool iommu_is_attach_deferred(struct device *dev)
>   
>   static int iommu_group_do_dma_first_attach(struct iommu_group *group, struct device *dev)
>   {
> -	return __iommu_device_set_domain(group, dev, group->domain,
> -					 IOMMU_SET_DOMAIN_WITH_DEFERRED);
> +	return __iommu_device_set_domain(
> +		group, dev, group->domain,
> +		group->owner ? 0 : IOMMU_SET_DOMAIN_WITH_DEFERRED);
>   }
>   
>   int iommu_probe_device(struct device *dev)

  reply	other threads:[~2023-04-06 16:38 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-04-05 23:38 [PATCH v3 00/17] Consolidate the error handling around device attachment Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 01/17] iommu: Replace iommu_group_device_count() with list_count_nodes() Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 02/17] iommu: Add for_each_group_device() Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 03/17] iommu: Make __iommu_group_set_domain() handle error unwind Jason Gunthorpe
2023-04-06 18:08   ` Robin Murphy
2023-04-06 19:09     ` Jason Gunthorpe
2023-04-13 16:47       ` Robin Murphy
2023-04-13 18:39         ` Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 04/17] iommu: Use __iommu_group_set_domain() for __iommu_attach_group() Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 05/17] iommu: Use __iommu_group_set_domain() in iommu_change_dev_def_domain() Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 06/17] iommu: Replace __iommu_group_dma_first_attach() with set_domain Jason Gunthorpe
2023-04-06 15:58   ` Robin Murphy
2023-04-06 19:17     ` Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 07/17] iommu: Make iommu_group_do_dma_first_attach() simpler Jason Gunthorpe
2023-04-06  3:28   ` Baolu Lu
2023-04-06 15:44   ` Robin Murphy
2023-04-06 16:09     ` Jason Gunthorpe
2023-04-06 19:05       ` Robin Murphy
2023-04-05 23:38 ` [PATCH v3 08/17] iommu: Make iommu_group_do_dma_first_attach() work with owned groups Jason Gunthorpe
2023-04-06 16:37   ` Robin Murphy [this message]
2023-04-06 19:34     ` Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 09/17] iommu: Fix iommu_probe_device() to attach the right domain Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 10/17] iommu: Do iommu_group_create_direct_mappings() before attach Jason Gunthorpe
2023-04-06  3:14   ` Baolu Lu
2023-04-05 23:38 ` [PATCH v3 11/17] iommu: Remove the assignment of group->domain during default domain alloc Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 12/17] iommu: Consolidate the code to calculate the target default domain type Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 13/17] iommu: Revise iommu_group_alloc_default_domain() Jason Gunthorpe
2023-04-06  3:40   ` Baolu Lu
2023-04-06 11:34     ` Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 14/17] iommu: Consolidate the default_domain setup to one function Jason Gunthorpe
2023-04-06  4:43   ` Baolu Lu
2023-04-05 23:38 ` [PATCH v3 15/17] iommu: Allow IOMMU_RESV_DIRECT to work on ARM Jason Gunthorpe
2023-04-06  3:53   ` Baolu Lu
2023-04-05 23:38 ` [PATCH v3 16/17] iommu: Remove __iommu_group_for_each_dev() Jason Gunthorpe
2023-04-05 23:38 ` [PATCH v3 17/17] iommu: Tidy the control flow in iommu_group_store_type() Jason Gunthorpe
2023-04-07 11:02 ` [PATCH v3 00/17] Consolidate the error handling around device attachment Naresh Kamboju

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=db92f18c-e2ed-78d0-db7e-83d8ee8a4dce@arm.com \
    --to=robin.murphy@arm.com \
    --cc=baolu.lu@linux.intel.com \
    --cc=iommu@lists.linux.dev \
    --cc=jgg@nvidia.com \
    --cc=joro@8bytes.org \
    --cc=kevin.tian@intel.com \
    --cc=llvm@lists.linux.dev \
    --cc=nathan@kernel.org \
    --cc=ndesaulniers@google.com \
    --cc=nicolinc@nvidia.com \
    --cc=ojeda@kernel.org \
    --cc=trix@redhat.com \
    --cc=will@kernel.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.