Re: [dpdk-dev] [PATCH] pci: keep API compatibility with mmap values

[Thomas Monjalon <thomas@monjalon.net>]

10/07/2020 17:39, Burakov, Anatoly:
> On 10-Jul-20 12:53 PM, Thomas Monjalon wrote:
> > The function pci_map_resource() returns MAP_FAILED in case of error.
> > When replacing the call to mmap() by rte_mem_map(),
> > the error code became NULL, breaking the API.
> > This function is probably not used outside of DPDK,
> > but it is still a problem for two reasons:
> > 	- the deprecation process was not followed
> > 	- the Linux function pci_vfio_mmap_bar() is broken for i40e
> > 
> > The error code is reverted to the Unix value MAP_FAILED.
> > Windows needs to define this special value (-1 as in Unix).
> > After proper deprecation process, the API could be changed again
> > if really needed.
> > 
> > Because of the switch from mmap() to rte_mem_map(),
> > another part of the API was changed: "int additional_flags"
> > are defined as "additional flags for the mapping range"
> > without mentioning it was directly used in mmap().
> > Currently it is directly used in rte_mem_map(),
> > that's why the values rte_map_flags must be mapped (sic) on the mmap ones
> > in case of Unix OS.
> > 
> > These are side effects of a badly defined API using Unix values.
[...]
> >   /** Additional flags for memory mapping. */
> >   enum rte_map_flags {
> > +#ifdef RTE_EXEC_ENV_WINDOWS
> >   	/** Changes to the mapped memory are visible to other processes. */
> >   	RTE_MAP_SHARED = 1 << 0,
> >   	/** Mapping is not backed by a regular file. */
> > @@ -35,6 +37,12 @@ enum rte_map_flags {
> >   	 * it is not required to do so, thus mapping with this flag may fail.
> >   	 */
> >   	RTE_MAP_FORCE_ADDRESS = 1 << 3
> > +#else /* map mmap flags because they are exposed in pci_map_resource() API */
> > +	RTE_MAP_SHARED = MAP_SHARED,
> > +	RTE_MAP_ANONYMOUS = MAP_ANONYMOUS,
> > +	RTE_MAP_PRIVATE = MAP_PRIVATE,
> > +	RTE_MAP_FORCE_ADDRESS = MAP_FIXED,
> > +#endif
> 
> I'm probably missing something, but why is this needed? Doesn't 

Yes you missed reading the commit log :)
Or maybe it is not written clearly enough. Will try to rephrase.

> rte_mem_map() automatically translate these flags into proper ones? 
> pci_map_resource() will call rte_mem_map(), and that will translate 
> these flags into their Unix equivalents.

The problem is that we have an API which is taking mmap flags as input.
"int additional_flags" is a parameter of the function,
and are supposed to be mmap flags. But it is not stated clearly.
When Windows will use this function, it won't use mmap flags
but RTE_MAP_*. So we must accept both.
That's why the best is to make values the same.

In 20.11, we could change the API,
make clear that only RTE_MAP_* is accepted,
and remove this workaround.
Or even better, remove pci_map_resource from the PCI lib,
and implement it in the PCI bus driver.

pci_map_resource() function is a bad designed API