Received-SPF: pass (google.com: domain of linux-kernel+bounces-10151-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Message-ID: <59bb418f-e78d-4bad-b3d4-997f6b30993a@linux.intel.com>
Date: Sat, 23 Dec 2023 08:24:53 +0800
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [RFC PATCH v5 3/3] iommu/vt-d: abort the devTLB invalidation
 waiting if device is removed
From: Ethan Zhao <haifeng.zhao@linux.intel.com>
To: bhelgaas@google.com, baolu.lu@linux.intel.com, dwmw2@infradead.org,
 will@kernel.org, robin.murphy@arm.com, lukas@wunner.de
Cc: linux-pci@vger.kernel.org, iommu@lists.linux.dev,
 linux-kernel@vger.kernel.org
References: <20231222104108.18499-1-haifeng.zhao@linux.intel.com>
 <20231222104108.18499-4-haifeng.zhao@linux.intel.com>
 <07285ebd-ae7f-4c4a-a926-8c52a2da8da6@linux.intel.com>
In-Reply-To: <07285ebd-ae7f-4c4a-a926-8c52a2da8da6@linux.intel.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit


On 12/23/2023 7:35 AM, Ethan Zhao wrote:
>
> On 12/22/2023 6:41 PM, Ethan Zhao wrote:
>> Even the devTLB invalidation request is just submitted and waiting it
>> to be done/timeout in qi_submit_sync(), it is possible device is removed
>> or powered-off. try to break it out in such rare but possible case.
>>
>> This patch is sent for more comment. not tested, only passed compiling.
>>
>> Signed-off-by: Ethan Zhao <haifeng.zhao@linux.intel.com>
>> ---
>>   drivers/iommu/intel/dmar.c  |  3 ++-
>>   drivers/iommu/intel/iommu.c | 36 ++++++++++++++++++++++++++++++++++++
>>   2 files changed, 38 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/iommu/intel/dmar.c b/drivers/iommu/intel/dmar.c
>> index 23cb80d62a9a..d8637ab93387 100644
>> --- a/drivers/iommu/intel/dmar.c
>> +++ b/drivers/iommu/intel/dmar.c
>> @@ -1422,7 +1422,8 @@ int qi_submit_sync(struct intel_iommu *iommu, 
>> struct qi_desc *desc,
>>        */
>>       writel(qi->free_head << shift, iommu->reg + DMAR_IQT_REG);
>>   -    while (qi->desc_status[wait_index] != QI_DONE) {
>> +    while (qi->desc_status[wait_index] != QI_DONE &&
>> +            qi->desc_status[wait_index] != QI_ABORT) {
>
> Another way is checking pci_device_is_present() here and bail out,
>
> how about it ?
>
>>           /*
>>            * We will leave the interrupts disabled, to prevent interrupt
>>            * context to queue another cmd while a cmd is already 
>> submitted
>> diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
>> index 897159dba47d..33075d0688bc 100644
>> --- a/drivers/iommu/intel/iommu.c
>> +++ b/drivers/iommu/intel/iommu.c
>> @@ -4472,10 +4472,46 @@ static struct iommu_device 
>> *intel_iommu_probe_device(struct device *dev)
>>       return &iommu->iommu;
>>   }
>>   +static void intel_iommu_abort_devtlib_invalidate(struct device *dev)
>> +{
>> +    struct device_domain_info *info = dev_iommu_priv_get(dev);
>> +    struct intel_iommu *iommu = info->iommu;
>> +    struct q_inval *qi = iommu->qi;
>> +    struct qi_desc *desc, *idesc;
>> +    int index, offset, shift;
>> +    u16 sid, qdep, pfsid
>> +    unsigned long flags;
>> +
>> +    if (!dev_is_pci(info->dev) || !info->ats_enabled || !qi)
>> +        return;
>> +    if (!pci_dev_is_disconnected(to_pci_dev(dev)))
>> +        return;
>> +
>> +    sid = info->bus << 8 | info->devfn;
>> +    qdep = info->ats_qdep;
>> +    pfsid = info->pfsid;
>> +
>> +    raw_spin_lock_irqsave(&qi->q_lock, flags);
>> +    for (index = 1; index < QI_LENGTH; index++) {
>> +        offset = index << shift;
>> +        desc = qi->desc + offset;
>> +        if (desc->qw0 & QI_IWD_TYPE) {
>> +            offset = (index-1) << shift;
>> +            idesc = qi->desc + offset;
>> +            if (idesc->qw0 & QI_DEV_EIOTLB_SID(sid)) {
>> +                if (qi->desc_status[index] == QI_IN_USE)
>> +                    qi->desc_status[index] = QI_ABORT;
>> +            }
>> +        }
>> +    }
>> +    raw_spin_unlock_irqrestore(&qi->q_lock, flags);
>> +
>> +}
>>   static void intel_iommu_release_device(struct device *dev)
>>   {
>>       struct device_domain_info *info = dev_iommu_priv_get(dev);
>>   +    intel_iommu_abort_devtlib_invalidate(dev);
>
> Wonder if there is lock something prevent pciehp_ist() supprise_removal

pci_lock_rescan_remove() will block another thread to re-enter the function

pci_stop_and_remove_bus_device() to get here, this patch doesn't work.


>
> interrupt response re-enter to get here when another safe_removal is 
> in process,
>
> if so ,  see above
>
>>       dmar_remove_one_dev_info(dev);
>>       intel_pasid_free_table(dev);
>>       intel_iommu_debugfs_remove_dev(info);
>