Re: [PATCH v7 01/14] nvme-core: introduce nvme_ctrl_get_by_path()

[Logan Gunthorpe <logang@deltatee.com>]

On 2019-08-15 5:46 a.m., Max Gurtovoy wrote:
> 
> On 8/2/2019 2:45 AM, Logan Gunthorpe wrote:
>> nvme_ctrl_get_by_path() is analagous to blkdev_get_by_path() except it
>> gets a struct nvme_ctrl from the path to its char dev (/dev/nvme0).
>> It makes use of filp_open() to open the file and uses the private
>> data to obtain a pointer to the struct nvme_ctrl. If the fops of the
>> file do not match, -EINVAL is returned.
>>
>> The purpose of this function is to support NVMe-OF target passthru.
>>
>> Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
>> ---
>>   drivers/nvme/host/core.c | 24 ++++++++++++++++++++++++
>>   drivers/nvme/host/nvme.h |  2 ++
>>   2 files changed, 26 insertions(+)
>>
>> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
>> index e6ee6f2a3da6..f72334f34a30 100644
>> --- a/drivers/nvme/host/core.c
>> +++ b/drivers/nvme/host/core.c
>> @@ -2817,6 +2817,30 @@ static const struct file_operations
>> nvme_dev_fops = {
>>       .compat_ioctl    = nvme_dev_ioctl,
>>   };
>>   +struct nvme_ctrl *nvme_ctrl_get_by_path(const char *path)
>> +{
>> +    struct nvme_ctrl *ctrl;
>> +    struct file *f;
>> +
>> +    f = filp_open(path, O_RDWR, 0);
>> +    if (IS_ERR(f))
>> +        return ERR_CAST(f);
>> +
>> +    if (f->f_op != &nvme_dev_fops) {
>> +        ctrl = ERR_PTR(-EINVAL);
>> +        goto out_close;
>> +    }
> 
> Logan,
> 
> this means that the PT is for nvme-pci and also nvme-fabrics as well.
> 
> Is this the intention ? or we want to restrict it to pci only.

Yes, in theory, someone could passthru an nvme-fabrics controller or
they could passthru a passthru'd passthru'd nvme-fabrics controller.
This probably isn't a good idea but I don't know that we need to
specifically reject it. If you think we should I could figure out a way
to filter by pci controllers only.

Logan