Re: [PATCH v3 1/2] uio_hv_generic: Fix sysfs creation path for ring buffer

[Naman Jain <namjain@linux.microsoft.com>]

On 3/31/2025 11:08 AM, Naman Jain wrote:
> 
> 
> On 3/30/2025 9:05 PM, Michael Kelley wrote:
>> From: Naman Jain <namjain@linux.microsoft.com> Sent: Thursday, March 
>> 27, 2025 10:28 PM
>>>
>>> On regular bootup, devices get registered to VMBus first, so when
>>> uio_hv_generic driver for a particular device type is probed,
>>> the device is already initialized and added, so sysfs creation in
>>> uio_hv_generic probe works fine. However, when device is removed
>>> and brought back, the channel rescinds and device again gets
>>> registered to VMBus. However this time, the uio_hv_generic driver is
>>> already registered to probe for that device and in this case sysfs
>>> creation is tried before the device's kobject gets initialized
>>> completely.
>>>
>>> Fix this by moving the core logic of sysfs creation for ring buffer,
>>> from uio_hv_generic to HyperV's VMBus driver, where rest of the sysfs
>>> attributes for the channels are defined. While doing that, make use
>>> of attribute groups and macros, instead of creating sysfs directly,
>>> to ensure better error handling and code flow.
>>>
>>> Problem path:
>>> vmbus_process_offer (new offer comes for the VMBus device)
>>>    vmbus_add_channel_work
>>>      vmbus_device_register
>>>        |-> device_register
>>>        |     |...
>>>        |     |-> hv_uio_probe
>>>        |           |...
>>>        |           |-> sysfs_create_bin_file (leads to a warning as
>>>        |                 primary channel's kobject, which is used to
>>>        |                 create sysfs is not yet initialized)
>>>        |-> kset_create_and_add
>>>        |-> vmbus_add_channel_kobj (initialization of primary channel's
>>>                                    kobject happens later)
>>>
>>> Above code flow is sequential and the warning is always reproducible in
>>> this path.
>>>
>>> Fixes: 9ab877a6ccf8 ("uio_hv_generic: make ring buffer attribute for 
>>> primary channel")
>>> Cc: stable@kernel.org
>>> Suggested-by: Saurabh Sengar <ssengar@linux.microsoft.com>
>>> Suggested-by: Michael Kelley <mhklinux@outlook.com>
>>> Signed-off-by: Naman Jain <namjain@linux.microsoft.com>
>>> ---
>>>   drivers/hv/hyperv_vmbus.h    |   6 ++
>>>   drivers/hv/vmbus_drv.c       | 110 ++++++++++++++++++++++++++++++++++-
>>>   drivers/uio/uio_hv_generic.c |  33 +++++------
>>>   include/linux/hyperv.h       |   6 ++
>>>   4 files changed, 134 insertions(+), 21 deletions(-)
>>>
>>
>> [snip]
>>
>>> +/**
>>> + * hv_create_ring_sysfs() - create "ring" sysfs entry corresponding 
>>> to ring buffers for a channel.
>>> + * @channel: Pointer to vmbus_channel structure
>>> + * @hv_mmap_ring_buffer: function pointer for initializing the 
>>> function to be called on mmap of
>>> + *                       channel's "ring" sysfs node, which is for 
>>> the ring buffer of that channel.
>>> + *                       Function pointer is of below type:
>>> + *                       int (*hv_mmap_ring_buffer)(struct 
>>> vmbus_channel *channel,
>>> + *                                                  struct 
>>> vm_area_struct *vma))
>>> + *                       This has a pointer to the channel and a 
>>> pointer to vm_area_struct,
>>> + *                       used for mmap, as arguments.
>>> + *
>>> + * Sysfs node for ring buffer of a channel is created along with 
>>> other fields, however its
>>> + * visibility is disabled by default. Sysfs creation needs to be 
>>> controlled when the use-case
>>> + * is running.
>>> + * For example, HV_NIC device is used either by uio_hv_generic or 
>>> hv_netvsc at any given point of
>>> + * time, and "ring" sysfs is needed only when uio_hv_generic is 
>>> bound to that device. To avoid
>>> + * exposing the ring buffer by default, this function is reponsible 
>>> to enable visibility of
>>> + * ring for userspace to use.
>>> + * Note: Race conditions can happen with userspace and it is not 
>>> encouraged to create new
>>> + * use-cases for this. This was added to maintain backward 
>>> compatibility, while solving
>>> + * one of the race conditions in uio_hv_generic while creating sysfs.
>>> + *
>>> + * Returns 0 on success or error code on failure.
>>> + */
>>> +int hv_create_ring_sysfs(struct vmbus_channel *channel,
>>> +             int (*hv_mmap_ring_buffer)(struct vmbus_channel *channel,
>>> +                            struct vm_area_struct *vma))
>>> +{
>>> +    struct kobject *kobj = &channel->kobj;
>>> +    struct vmbus_channel *primary_channel = channel->primary_channel ?
>>> +        channel->primary_channel : channel;
>>> +
>>> +    channel->mmap_ring_buffer = hv_mmap_ring_buffer;
>>> +    channel->ring_sysfs_visible = true;
>>> +
>>> +    /*
>>> +     * Skip updating the sysfs group if the primary channel is not 
>>> yet initialized and sysfs
>>> +     * group is not yet created. In those cases, the 'ring' will be 
>>> created later in
>>> +     * vmbus_device_register() -> vmbus_add_channel_kobj().
>>> +     */
>>> +    if  (!primary_channel->device_obj->channels_kset)
>>> +        return 0;
>>
>> This test doesn't accomplish what you want. It tests if the "channels" 
>> directory
>> has been created, but not if the numbered subdirectory for this 
>> channel has been
>> created. sysfs_update_group() operates on the numbered subdirectory and
>> could still fail because it hasn't been created yet.
>>
>> My recommendation is to not try to do a test, and just let 
>> sysfs_update_group()
>> fail in that case (and ignore the error).
>>
>> Michael
>>
> 
> Thanks Michael. Will remove it.
> 
> Regards,
> Naman

Just to let you know, I'll wait till the end of merge window and a few 
more days before sending the next version. Meanwhile if there are any 
further comments, I'll take them up together.

Regards,
Naman

> 
>>> +
>>> +    return sysfs_update_group(kobj, &vmbus_chan_group);
>>> +}
>>> +EXPORT_SYMBOL_GPL(hv_create_ring_sysfs);
>