Re: Re: [PATCH] iommu/arm-smmu: Fix arm_smmu_device refcount leak when arm_smmu_rpm_get fails

[Xiyu Yang]

Thanks for your advice, I'll send a v2 patch soon.


> -----Original Messages-----
> From: "Robin Murphy" <robin.murphy@xxxxxxx>
> Sent Time: 2021-06-09 22:12:11 (Wednesday)
> To: "Xiyu Yang" <xiyuyang19@xxxxxxxxxxxx>, "Will Deacon" <will@xxxxxxxxxx>, "Joerg Roedel" <joro@xxxxxxxxxx>, "Nicolin Chen" <nicoleotsuka@xxxxxxxxx>, "Bjorn Andersson" <bjorn.andersson@xxxxxxxxxx>, "Krishna Reddy" <vdumpa@xxxxxxxxxx>, "Jordan Crouse" <jordan@xxxxxxxxxxxxxxxxx>, "Sai Prakash Ranjan" <saiprakash.ranjan@xxxxxxxxxxxxxx>, linux-arm-kernel@xxxxxxxxxxxxxxxxxxx, iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx
> Cc: yuanxzhang@xxxxxxxxxxxx, "Xin Tan" <tanxin.ctf@xxxxxxxxx>
> Subject: Re: [PATCH] iommu/arm-smmu: Fix arm_smmu_device refcount leak when arm_smmu_rpm_get fails
> 
> On 2021-06-09 14:35, Xiyu Yang wrote:
> > arm_smmu_rpm_get() invokes pm_runtime_get_sync(), which increases the
> > refcount of the "smmu" even though the return value is less than 0.
> > 
> > The reference counting issue happens in some error handling paths of
> > arm_smmu_rpm_get() in its caller functions. When arm_smmu_rpm_get()
> > fails, the caller functions forget to decrease the refcount of "smmu"
> > increased by arm_smmu_rpm_get(), causing a refcount leak.
> > 
> > Fix this issue by calling arm_smmu_rpm_put() or jumping to the "rpm_put"
> > label when arm_smmu_rpm_get() fails.
> 
> If only there was some kind of helper function which could encapsulate 
> the correct expected behaviour in a single place...
> 
> In fact with the new pm_runtime_resume_and_get() API I think these two 
> patches boil down to a one-line change.
> 
> Thanks,
> Robin.
> 
> > Signed-off-by: Xiyu Yang <xiyuyang19@xxxxxxxxxxxx>
> > Signed-off-by: Xin Tan <tanxin.ctf@xxxxxxxxx>
> > ---
> >   drivers/iommu/arm/arm-smmu/arm-smmu.c | 13 +++++++++----
> >   1 file changed, 9 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > index 6f72c4d208ca..177ee54c5534 100644
> > --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > @@ -840,7 +840,7 @@ static void arm_smmu_destroy_domain_context(struct iommu_domain *domain)
> >   
> >   	ret = arm_smmu_rpm_get(smmu);
> >   	if (ret < 0)
> > -		return;
> > +		goto rpm_put;
> >   
> >   	/*
> >   	 * Disable the context bank and free the page tables before freeing
> > @@ -857,6 +857,7 @@ static void arm_smmu_destroy_domain_context(struct iommu_domain *domain)
> >   	free_io_pgtable_ops(smmu_domain->pgtbl_ops);
> >   	__arm_smmu_free_bitmap(smmu->context_map, cfg->cbndx);
> >   
> > +rpm_put:
> >   	arm_smmu_rpm_put(smmu);
> >   }
> >   
> > @@ -1153,7 +1154,7 @@ static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)
> >   
> >   	ret = arm_smmu_rpm_get(smmu);
> >   	if (ret < 0)
> > -		return ret;
> > +		goto rpm_put;
> >   
> >   	/* Ensure that the domain is finalised */
> >   	ret = arm_smmu_init_domain_context(domain, smmu, dev);
> > @@ -1404,7 +1405,7 @@ static struct iommu_device *arm_smmu_probe_device(struct device *dev)
> >   
> >   	ret = arm_smmu_rpm_get(smmu);
> >   	if (ret < 0)
> > -		goto out_cfg_free;
> > +		goto rpm_put;
> >   
> >   	ret = arm_smmu_master_alloc_smes(dev);
> >   	arm_smmu_rpm_put(smmu);
> > @@ -1417,6 +1418,8 @@ static struct iommu_device *arm_smmu_probe_device(struct device *dev)
> >   
> >   	return &smmu->iommu;
> >   
> > +rpm_put:
> > +	arm_smmu_rpm_put(smmu);
> >   out_cfg_free:
> >   	kfree(cfg);
> >   out_free:
> > @@ -1438,8 +1441,10 @@ static void arm_smmu_release_device(struct device *dev)
> >   	smmu = cfg->smmu;
> >   
> >   	ret = arm_smmu_rpm_get(smmu);
> > -	if (ret < 0)
> > +	if (ret < 0) {
> > +		arm_smmu_rpm_put(smmu);
> >   		return;
> > +	}
> >   
> >   	arm_smmu_master_free_smes(cfg, fwspec);
> >   
> >