linux-kernel - Re: [PATCH 4/5] iommu/amd: Hold the domain lock when calling iommu_map

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <AC63DCD0-5322-4ECB-AB42-829AD4FBDFB9@amazon.de>
Date:   Fri, 20 Sep 2019 18:05:37 +0000
From:   "Sironi, Filippo" <sironi@...zon.de>
To:     "Sironi, Filippo" <sironi@...zon.de>,
        Joerg Roedel <joro@...tes.org>,
        Filippo Sironi via iommu <iommu@...ts.linux-foundation.org>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 4/5] iommu/amd: Hold the domain lock when calling
 iommu_map_page



> On 10. Sep 2019, at 19:49, Filippo Sironi <sironi@...zon.de> wrote:
> 
> iommu_map_page calls into __domain_flush_pages, which requires the
> domain lock since it traverses the device list, which the lock protects.
> 
> Signed-off-by: Filippo Sironi <sironi@...zon.de>
> ---
> drivers/iommu/amd_iommu.c | 5 +++++
> 1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
> index d4f25767622e..3714ae5ded31 100644
> --- a/drivers/iommu/amd_iommu.c
> +++ b/drivers/iommu/amd_iommu.c
> @@ -2562,6 +2562,7 @@ static int map_sg(struct device *dev, struct scatterlist *sglist,
> 	unsigned long address;
> 	u64 dma_mask;
> 	int ret;
> +	unsigned long flags;
> 
> 	domain = get_domain(dev);
> 	if (IS_ERR(domain))
> @@ -2587,7 +2588,9 @@ static int map_sg(struct device *dev, struct scatterlist *sglist,
> 
> 			bus_addr  = address + s->dma_address + (j << PAGE_SHIFT);
> 			phys_addr = (sg_phys(s) & PAGE_MASK) + (j << PAGE_SHIFT);
> +			spin_lock_irqsave(&domain->lock, flags);
> 			ret = iommu_map_page(domain, bus_addr, phys_addr, PAGE_SIZE, prot, GFP_ATOMIC);
> +			spin_unlock_irqrestore(&domain->lock, flags);
> 			if (ret)
> 				goto out_unmap;
> 
> @@ -3095,7 +3098,9 @@ static int amd_iommu_map(struct iommu_domain *dom, unsigned long iova,
> 		prot |= IOMMU_PROT_IW;
> 
> 	mutex_lock(&domain->api_lock);
> +	spin_lock(&domain->lock);
> 	ret = iommu_map_page(domain, iova, paddr, page_size, prot, GFP_KERNEL);
> +	spin_unlock(&domain->lock);
> 	mutex_unlock(&domain->api_lock);

The spin_lock/spin_unlock aren't the correct choice.
These should be spin_lock_irqsave and spin_unlock_irqrestore.
Of course, with the variant Joerg suggested, this isn't a
problem anymore.

> 	domain_flush_np_cache(domain, iova, page_size);
> -- 
> 2.7.4
> 




Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879