Re: [PATCH v5 04/17] device-dax/kmem: replace release_resource() with release_mem_region()

[Dan Williams]

On Wed, Sep 30, 2020 at 9:23 AM David Hildenbrand <david@xxxxxxxxxx> wrote:
>
> On 25.09.20 21:12, Dan Williams wrote:
> > Towards removing the mode specific @dax_kmem_res attribute from the
> > generic 'struct dev_dax', and preparing for multi-range support, change
> > the kmem driver to use the idiomatic release_mem_region() to pair with
> > the initial request_mem_region(). This also eliminates the need to open
> > code the release of the resource allocated by request_mem_region().
> >
> > As there are no more dax_kmem_res users, delete this struct member.
> >
> > Cc: David Hildenbrand <david@xxxxxxxxxx>
> > Cc: Vishal Verma <vishal.l.verma@xxxxxxxxx>
> > Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
> > Cc: Pavel Tatashin <pasha.tatashin@xxxxxxxxxx>
> > Cc: Brice Goglin <Brice.Goglin@xxxxxxxx>
> > Cc: Dave Jiang <dave.jiang@xxxxxxxxx>
> > Cc: David Hildenbrand <david@xxxxxxxxxx>
> > Cc: Ira Weiny <ira.weiny@xxxxxxxxx>
> > Cc: Jia He <justin.he@xxxxxxx>
> > Cc: Joao Martins <joao.m.martins@xxxxxxxxxx>
> > Cc: Jonathan Cameron <Jonathan.Cameron@xxxxxxxxxx>
> > Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
> > ---
> >  drivers/dax/dax-private.h |    3 ---
> >  drivers/dax/kmem.c        |   20 +++++++-------------
> >  2 files changed, 7 insertions(+), 16 deletions(-)
> >
> > diff --git a/drivers/dax/dax-private.h b/drivers/dax/dax-private.h
> > index 6779f683671d..12a2dbc43b40 100644
> > --- a/drivers/dax/dax-private.h
> > +++ b/drivers/dax/dax-private.h
> > @@ -42,8 +42,6 @@ struct dax_region {
> >   * @dev - device core
> >   * @pgmap - pgmap for memmap setup / lifetime (driver owned)
> >   * @range: resource range for the instance
> > - * @dax_mem_res: physical address range of hotadded DAX memory
> > - * @dax_mem_name: name for hotadded DAX memory via add_memory_driver_managed()
> >   */
> >  struct dev_dax {
> >       struct dax_region *region;
> > @@ -52,7 +50,6 @@ struct dev_dax {
> >       struct device dev;
> >       struct dev_pagemap *pgmap;
> >       struct range range;
> > -     struct resource *dax_kmem_res;
> >  };
> >
> >  static inline u64 range_len(struct range *range)
> > diff --git a/drivers/dax/kmem.c b/drivers/dax/kmem.c
> > index 6fe2cb1c5f7c..e56fc688bdc5 100644
> > --- a/drivers/dax/kmem.c
> > +++ b/drivers/dax/kmem.c
> > @@ -33,7 +33,7 @@ int dev_dax_kmem_probe(struct device *dev)
> >  {
> >       struct dev_dax *dev_dax = to_dev_dax(dev);
> >       struct range range = dax_kmem_range(dev_dax);
> > -     struct resource *new_res;
> > +     struct resource *res;
> >       char *res_name;
> >       int numa_node;
> >       int rc;
> > @@ -56,8 +56,8 @@ int dev_dax_kmem_probe(struct device *dev)
> >               return -ENOMEM;
> >
> >       /* Region is permanently reserved if hotremove fails. */
> > -     new_res = request_mem_region(range.start, range_len(&range), res_name);
> > -     if (!new_res) {
> > +     res = request_mem_region(range.start, range_len(&range), res_name);
> > +     if (!res) {
> >               dev_warn(dev, "could not reserve region [%#llx-%#llx]\n", range.start, range.end);
> >               kfree(res_name);
> >               return -EBUSY;
> > @@ -69,23 +69,20 @@ int dev_dax_kmem_probe(struct device *dev)
> >        * inherit flags from the parent since it may set new flags
> >        * unknown to us that will break add_memory() below.
> >        */
> > -     new_res->flags = IORESOURCE_SYSTEM_RAM;
> > +     res->flags = IORESOURCE_SYSTEM_RAM;
> >
> >       /*
> >        * Ensure that future kexec'd kernels will not treat this as RAM
> >        * automatically.
> >        */
> > -     rc = add_memory_driver_managed(numa_node, new_res->start,
> > -                                    resource_size(new_res), kmem_name);
> > +     rc = add_memory_driver_managed(numa_node, range.start, range_len(&range), kmem_name);
> >       if (rc) {
> > -             release_resource(new_res);
> > -             kfree(new_res);
> > +             release_mem_region(range.start, range_len(&range));
> >               kfree(res_name);
> >               return rc;
> >       }
> >
> >       dev_set_drvdata(dev, res_name);
> > -     dev_dax->dax_kmem_res = new_res;
> >
> >       return 0;
> >  }
> > @@ -95,7 +92,6 @@ static int dev_dax_kmem_remove(struct device *dev)
> >  {
> >       struct dev_dax *dev_dax = to_dev_dax(dev);
> >       struct range range = dax_kmem_range(dev_dax);
> > -     struct resource *res = dev_dax->dax_kmem_res;
> >       const char *res_name = dev_get_drvdata(dev);
> >       int rc;
> >
> > @@ -114,10 +110,8 @@ static int dev_dax_kmem_remove(struct device *dev)
> >       }
> >
> >       /* Release and free dax resources */
> > -     release_resource(res);
> > -     kfree(res);
> > +     release_mem_region(range.start, range_len(&range));
>
> Does that work? AFAIKs,
>
> __release_region(&iomem_resource, (start), (n)) -> __release_region()
>
> will only remove stuff that is IORESOURCE_BUSY.

Oh, crud, indeed. More fallout from futzing with the flags.

> Maybe storing it in drvdata is indeed the easiest way to remove it from
> struct dax_region.

My first choice is still to get the driver out of the layering
violation of touching resource flags, but we don't seem to be
converging on a way forward there. So, stuffing more tracking into
drvdata (especially for the multi-range case) is the next best option.