linuxppc-dev.lists.ozlabs.org archive mirror
 help / color / mirror / Atom feed
From: "Janne Grunau" <j@jannau.net>
To: "Jason Gunthorpe" <jgg@nvidia.com>,
	"Alyssa Rosenzweig" <alyssa@rosenzweig.io>,
	asahi@lists.linux.dev,
	"Christophe Leroy" <christophe.leroy@csgroup.eu>,
	"David Woodhouse" <dwmw2@infradead.org>,
	iommu@lists.linux.dev, "Joerg Roedel" <joro@8bytes.org>,
	"Kevin Tian" <kevin.tian@intel.com>,
	linux-arm-kernel@lists.infradead.org,
	linuxppc-dev@lists.ozlabs.org, "Hector Martin" <marcan@marcan.st>,
	"Michael Ellerman" <mpe@ellerman.id.au>,
	"Nicholas Piggin" <npiggin@gmail.com>,
	"Robin Murphy" <robin.murphy@arm.com>,
	"Sven Peter" <sven@svenpeter.dev>,
	"Will Deacon" <will@kernel.org>
Cc: Lu Baolu <baolu.lu@linux.intel.com>
Subject: Re: [PATCH v2 9/9] iommu/dart: Remove the force_bypass variable
Date: Mon, 16 Oct 2023 13:44:29 +0200	[thread overview]
Message-ID: <2111fedf-a690-453a-a705-6e9b8141d93d@app.fastmail.com> (raw)
In-Reply-To: <9-v2-bff223cf6409+282-dart_paging_jgg@nvidia.com>

Hej,

On Thu, Sep 28, 2023, at 01:47, Jason Gunthorpe wrote:
> This flag just caches if the IO page size is larger than the CPU
> PAGE_SIZE. This only needs to be checked in two places so remove the
> confusingly named cache.
>
> dart would like to not support paging domains at all if the IO page size
> is larger than the CPU page size. In this case we should ideally fail
> domain_alloc_paging(), as there is no point in creating a domain that can
> never be attached. Move the test into apple_dart_finalize_domain().
>
> The check in apple_dart_mod_streams() will prevent the domain from being
> attached to the wrong dart
>
> There is no HW limitation that prevents BLOCKED domains from working,
> remove that test.
>
> The check in apple_dart_of_xlate() is redundant since immediately after
> the pgsize is checked. Remove it.
>
> Remove the variable.
>
> Suggested-by: Janne Grunau <j@jannau.net>
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> ---
>  drivers/iommu/apple-dart.c | 20 ++++++--------------
>  1 file changed, 6 insertions(+), 14 deletions(-)
>
> diff --git a/drivers/iommu/apple-dart.c b/drivers/iommu/apple-dart.c
> index 126da0d89f0dd4..821b4a3465dfb9 100644
> --- a/drivers/iommu/apple-dart.c
> +++ b/drivers/iommu/apple-dart.c
> @@ -196,7 +196,6 @@ struct apple_dart_hw {
>   * @lock: lock for hardware operations involving this dart
>   * @pgsize: pagesize supported by this DART
>   * @supports_bypass: indicates if this DART supports bypass mode
> - * @force_bypass: force bypass mode due to pagesize mismatch?
>   * @sid2group: maps stream ids to iommu_groups
>   * @iommu: iommu core device
>   */
> @@ -217,7 +216,6 @@ struct apple_dart {
>  	u32 pgsize;
>  	u32 num_streams;
>  	u32 supports_bypass : 1;
> -	u32 force_bypass : 1;
> 
>  	struct iommu_group *sid2group[DART_MAX_STREAMS];
>  	struct iommu_device iommu;
> @@ -576,6 +574,9 @@ static int apple_dart_finalize_domain(struct 
> apple_dart_domain *dart_domain,
>  	int ret = 0;
>  	int i, j;
> 
> +	if (dart->pgsize > PAGE_SIZE)
> +		return -EINVAL;
> +
>  	mutex_lock(&dart_domain->init_lock);
> 
>  	if (dart_domain->finalized)
> @@ -659,9 +660,6 @@ static int apple_dart_attach_dev_paging(struct 
> iommu_domain *domain,
>  	struct apple_dart_master_cfg *cfg = dev_iommu_priv_get(dev);
>  	struct apple_dart_domain *dart_domain = to_dart_domain(domain);
> 
> -	if (cfg->stream_maps[0].dart->force_bypass)
> -		return -EINVAL;
> -
>  	ret = apple_dart_finalize_domain(dart_domain, cfg);
>  	if (ret)
>  		return ret;
> @@ -706,9 +704,6 @@ static int apple_dart_attach_dev_blocked(struct 
> iommu_domain *domain,
>  	struct apple_dart_stream_map *stream_map;
>  	int i;
> 
> -	if (cfg->stream_maps[0].dart->force_bypass)
> -		return -EINVAL;
> -
>  	for_each_stream_map(i, cfg, stream_map)
>  		apple_dart_hw_disable_dma(stream_map);
>  	return 0;
> @@ -803,8 +798,6 @@ static int apple_dart_of_xlate(struct device *dev, 
> struct of_phandle_args *args)
>  	if (cfg_dart) {
>  		if (cfg_dart->supports_bypass != dart->supports_bypass)
>  			return -EINVAL;
> -		if (cfg_dart->force_bypass != dart->force_bypass)
> -			return -EINVAL;
>  		if (cfg_dart->pgsize != dart->pgsize)
>  			return -EINVAL;
>  	}
> @@ -946,7 +939,7 @@ static int apple_dart_def_domain_type(struct device 
> *dev)
>  {
>  	struct apple_dart_master_cfg *cfg = dev_iommu_priv_get(dev);
> 
> -	if (cfg->stream_maps[0].dart->force_bypass)
> +	if (cfg->stream_maps[0].dart->pgsize > PAGE_SIZE)
>  		return IOMMU_DOMAIN_IDENTITY;
>  	if (!cfg->stream_maps[0].dart->supports_bypass)
>  		return IOMMU_DOMAIN_DMA;
> @@ -1146,8 +1139,6 @@ static int apple_dart_probe(struct platform_device *pdev)
>  		goto err_clk_disable;
>  	}
> 
> -	dart->force_bypass = dart->pgsize > PAGE_SIZE;
> -
>  	ret = apple_dart_hw_reset(dart);
>  	if (ret)
>  		goto err_clk_disable;
> @@ -1171,7 +1162,8 @@ static int apple_dart_probe(struct 
> platform_device *pdev)
>  	dev_info(
>  		&pdev->dev,
>  		"DART [pagesize %x, %d streams, bypass support: %d, bypass forced: 
> %d] initialized\n",
> -		dart->pgsize, dart->num_streams, dart->supports_bypass, 
> dart->force_bypass);
> +		dart->pgsize, dart->num_streams, dart->supports_bypass,
> +		dart->pgsize > PAGE_SIZE);
>  	return 0;
> 
>  err_sysfs_remove:
> -- 

Reviewed-by: Janne Grunau <j@jannau.net>

thanks,
Janne

  reply	other threads:[~2023-10-16 11:51 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-09-27 23:47 [PATCH v2 0/9] iommu: Convert dart & iommufd to the new domain_alloc_paging() Jason Gunthorpe
2023-09-27 23:47 ` [PATCH v2 1/9] iommu: Move IOMMU_DOMAIN_BLOCKED global statics to ops->blocked_domain Jason Gunthorpe
2023-10-09  7:47   ` Tian, Kevin
2023-09-27 23:47 ` [PATCH v2 2/9] iommu/vt-d: Update the definition of the blocking domain Jason Gunthorpe
2023-10-09  7:47   ` Tian, Kevin
2023-09-27 23:47 ` [PATCH v2 3/9] iommu/vt-d: Use ops->blocked_domain Jason Gunthorpe
2023-10-09  7:47   ` Tian, Kevin
2023-09-27 23:47 ` [PATCH v2 4/9] iommufd: Convert to alloc_domain_paging() Jason Gunthorpe
2023-10-09  7:48   ` Tian, Kevin
2023-09-27 23:47 ` [PATCH v2 5/9] iommu/dart: Use static global identity domains Jason Gunthorpe
2023-09-27 23:47 ` [PATCH v2 6/9] iommu/dart: Move the blocked domain support to a global static Jason Gunthorpe
2023-09-27 23:47 ` [PATCH v2 7/9] iommu/dart: Convert to domain_alloc_paging() Jason Gunthorpe
2023-09-27 23:47 ` [PATCH v2 8/9] iommu/dart: Call apple_dart_finalize_domain() as part of alloc_paging() Jason Gunthorpe
2023-09-27 23:47 ` [PATCH v2 9/9] iommu/dart: Remove the force_bypass variable Jason Gunthorpe
2023-10-16 11:44   ` Janne Grunau [this message]
2023-10-25 15:58 ` [PATCH v2 0/9] iommu: Convert dart & iommufd to the new domain_alloc_paging() jgg
2023-10-26  7:49 ` Joerg Roedel
2023-10-26 10:34   ` Sven Peter
2023-10-26 14:55     ` Joerg Roedel

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=2111fedf-a690-453a-a705-6e9b8141d93d@app.fastmail.com \
    --to=j@jannau.net \
    --cc=alyssa@rosenzweig.io \
    --cc=asahi@lists.linux.dev \
    --cc=baolu.lu@linux.intel.com \
    --cc=christophe.leroy@csgroup.eu \
    --cc=dwmw2@infradead.org \
    --cc=iommu@lists.linux.dev \
    --cc=jgg@nvidia.com \
    --cc=joro@8bytes.org \
    --cc=kevin.tian@intel.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linuxppc-dev@lists.ozlabs.org \
    --cc=marcan@marcan.st \
    --cc=mpe@ellerman.id.au \
    --cc=npiggin@gmail.com \
    --cc=robin.murphy@arm.com \
    --cc=sven@svenpeter.dev \
    --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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).