Re: [bug report] iommufd: Add a selftest

[Jason Gunthorpe <jgg@ziepe.ca>]

On Tue, Nov 15, 2022 at 03:49:44PM +0300, Dan Carpenter wrote:
> Hello Jason Gunthorpe,
> 
> The patch 147393321d65: "iommufd: Add a selftest" from Nov 19, 2021,
> leads to the following Smatch static checker warning:
> 
> 	drivers/iommu/iommufd/selftest.c:160 mock_domain_map_pages()
> 	warn: use 'gfp' here instead of GFP_XXX?
> 
> drivers/iommu/iommufd/selftest.c
>    133  static int mock_domain_map_pages(struct iommu_domain *domain,
>    134                                   unsigned long iova, phys_addr_t paddr,
>    135                                   size_t pgsize, size_t pgcount, int prot,
>    136                                   gfp_t gfp, size_t *mapped)
>    137  {
>    138          struct mock_iommu_domain *mock =
>    139                  container_of(domain, struct mock_iommu_domain, domain);
>    140          unsigned long flags = MOCK_PFN_START_IOVA;
>    141          unsigned long start_iova = iova;
>    142  
>    143          /*
>    144           * xarray does not reliably work with fault injection because it does a
>    145           * retry allocation, so put our own failure point.
>    146           */
>    147          if (iommufd_should_fail())
>    148                  return -ENOENT;
>    149  
>    150          WARN_ON(iova % MOCK_IO_PAGE_SIZE);
>    151          WARN_ON(pgsize % MOCK_IO_PAGE_SIZE);
>    152          for (; pgcount; pgcount--) {
>    153                  size_t cur;
>    154  
>    155                  for (cur = 0; cur != pgsize; cur += MOCK_IO_PAGE_SIZE) {
>    156                          void *old;
>    157  
>    158                          if (pgcount == 1 && cur + MOCK_IO_PAGE_SIZE == pgsize)
>    159                                  flags = MOCK_PFN_LAST_IOVA;
>    160                          old = xa_store(&mock->pfns, iova / MOCK_IO_PAGE_SIZE,
>    161                                         xa_mk_value((paddr / MOCK_IO_PAGE_SIZE) |
>    162                                                     flags),
>    163                                         GFP_KERNEL);
>                                                ^^^^^^^^^^
> "gfp" isn't used anywhere so maybe it was intended here?

It is always GFP_KERNEL, but yes it can use that. I fixed it

Jason