svn commit: r335094 - head/sys/ofed/drivers/infiniband/core

[Hans Petter Selasky]

On 06/14/18 02:03, Matthew Macy wrote:
> On Wed, Jun 13, 2018 at 4:47 PM, Ryan Libby <rlibby at freebsd.org> wrote:
>> On Wed, Jun 13, 2018 at 4:30 PM, Matt Macy <mmacy at freebsd.org> wrote:
>>> Author: mmacy
>>> Date: Wed Jun 13 23:30:54 2018
>>> New Revision: 335094
>>> URL: https://svnweb.freebsd.org/changeset/base/335094
>>>
>>> Log:
>>>    fix OFED build after r335053
>>>
>>> Modified:
>>>    head/sys/ofed/drivers/infiniband/core/ib_user_mad.c
>>>
>>> Modified: head/sys/ofed/drivers/infiniband/core/ib_user_mad.c
>>> ==============================================================================
>>> --- head/sys/ofed/drivers/infiniband/core/ib_user_mad.c Wed Jun 13 23:19:54 2018        (r335093)
>>> +++ head/sys/ofed/drivers/infiniband/core/ib_user_mad.c Wed Jun 13 23:30:54 2018        (r335094)
>>> @@ -130,7 +130,8 @@ struct ib_umad_packet {
>>>
>>>   static struct class *umad_class;
>>>
>>> -static const dev_t base_dev = MKDEV(IB_UMAD_MAJOR, IB_UMAD_MINOR_BASE);
>>> +#define IBMKDEV(x, y)  (((dev_t)(x) << 32) | (unsigned)(y))
>>> +static const dev_t base_dev = IBMKDEV(IB_UMAD_MAJOR, IB_UMAD_MINOR_BASE);
>>>
>>>   static DEFINE_SPINLOCK(port_lock);
>>>   static DECLARE_BITMAP(dev_map, IB_UMAD_MAX_PORTS);
>>>
>>
>> The scheme for major/minor encoding is different as of r335053.  Won't
>> that matter?
> 
> Yes.
> 
>> In sys/ofed/drivers/infiniband/core/{ib_ucm.c,ib_uverbs_main.c} the
>> pattern is to #define the MKDEV().  Following that would in
>> ib_user_mad.c would also resolve this.  Or makedev could be
>> re-macroized with the new scheme.
> 
> That's probably the best course of action.
> Will follow up.
> 
> 

Hi,

Please revert this patch and fix MKDEV() in:
compat/linuxkpi/common/include/linux/kdev_t.h

Instead.

The value returned by MKDEV() must be passable through MAJOR() and 
MINOR() to restore the two arguments. Please make sure this is the case. 
Else the code gets broken.

--HPS