Message ID | 20210928101521.3956331-1-aik@ozlabs.ru (mailing list archive) |
---|---|
State | Superseded |
Headers | show |
Series | [kernel] powerps/pseries/dma: Add support for 2M IOMMU page size | expand |
Related | show |
Context | Check | Description |
---|---|---|
snowpatch_ozlabs/github-powerpc_ppctests | success | Successfully ran 8 jobs. |
snowpatch_ozlabs/github-powerpc_selftests | success | Successfully ran 8 jobs. |
snowpatch_ozlabs/github-powerpc_sparse | success | Successfully ran 4 jobs. |
snowpatch_ozlabs/github-powerpc_kernel_qemu | success | Successfully ran 24 jobs. |
snowpatch_ozlabs/github-powerpc_clang | success | Successfully ran 7 jobs. |
Hello Alexey, On Tue, 2021-09-28 at 20:15 +1000, Alexey Kardashevskiy wrote: > The upcoming PAPR spec adds a 2M page size, bit 23 right after the 16G > page > size in the "ibm,query-pe-dma-window" call. > > This adds support for the new page size. Since the new page size is out > of sorted order, this changes the loop to not assume that shift[] is > sorted. > > Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> > --- > > This might not work if PHYP keeps rejecting new window requests for > less > than 32768 TCEs. This is needed: > https://github.com/aik/linux/commit/8cc8fa5ba5b3b4a18efbc9d81d9e5b85ca7c8a95 > > > --- > arch/powerpc/platforms/pseries/iommu.c | 10 +++++----- > 1 file changed, 5 insertions(+), 5 deletions(-) > > diff --git a/arch/powerpc/platforms/pseries/iommu.c > b/arch/powerpc/platforms/pseries/iommu.c > index c741689a5165..237bf405b0cb 100644 > --- a/arch/powerpc/platforms/pseries/iommu.c > +++ b/arch/powerpc/platforms/pseries/iommu.c > @@ -1159,14 +1159,15 @@ static void reset_dma_window(struct pci_dev > *dev, struct device_node *par_dn) > /* Return largest page shift based on "IO Page Sizes" output of > ibm,query-pe-dma-window. */ > static int iommu_get_page_shift(u32 query_page_size) > { > - /* Supported IO page-sizes according to LoPAR */ > + /* Supported IO page-sizes according to LoPAR, note that 2M is > out of order */ > const int shift[] = { > __builtin_ctzll(SZ_4K), __builtin_ctzll(SZ_64K), > __builtin_ctzll(SZ_16M), > __builtin_ctzll(SZ_32M), __builtin_ctzll(SZ_64M), > __builtin_ctzll(SZ_128M), > - __builtin_ctzll(SZ_256M), __builtin_ctzll(SZ_16G) > + __builtin_ctzll(SZ_256M), __builtin_ctzll(SZ_16G), > __builtin_ctzll(SZ_2M) > }; > > int i = ARRAY_SIZE(shift) - 1; > + int ret = 0; > > /* > * On LoPAR, ibm,query-pe-dma-window outputs "IO Page Sizes" > using a bit field: > @@ -1176,11 +1177,10 @@ static int iommu_get_page_shift(u32 > query_page_size) > */ > for (; i >= 0 ; i--) { > if (query_page_size & (1 << i)) > - return shift[i]; > + ret = max(ret, shift[i]); > } > > - /* No valid page size found. */ > - return 0; > + return ret; > } > > static struct property *ddw_property_create(const char *propname, u32 > liobn, u64 dma_addr, Looks great to me. FWIW: Reviewed-by: Leonardo Bras <leobras.c@gmail.com> Best regards, Leonardo
On 28/09/2021 12:15, Alexey Kardashevskiy wrote: > The upcoming PAPR spec adds a 2M page size, bit 23 right after the 16G page > size in the "ibm,query-pe-dma-window" call. > > This adds support for the new page size. Since the new page size is out > of sorted order, this changes the loop to not assume that shift[] is > sorted. > > Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> > --- Looks ok to me Reviewed-by: Frederic Barrat <fbarrat@linux.ibm.com> > > This might not work if PHYP keeps rejecting new window requests for less > than 32768 TCEs. This is needed: > https://github.com/aik/linux/commit/8cc8fa5ba5b3b4a18efbc9d81d9e5b85ca7c8a95 > > > --- > arch/powerpc/platforms/pseries/iommu.c | 10 +++++----- > 1 file changed, 5 insertions(+), 5 deletions(-) > > diff --git a/arch/powerpc/platforms/pseries/iommu.c b/arch/powerpc/platforms/pseries/iommu.c > index c741689a5165..237bf405b0cb 100644 > --- a/arch/powerpc/platforms/pseries/iommu.c > +++ b/arch/powerpc/platforms/pseries/iommu.c > @@ -1159,14 +1159,15 @@ static void reset_dma_window(struct pci_dev *dev, struct device_node *par_dn) > /* Return largest page shift based on "IO Page Sizes" output of ibm,query-pe-dma-window. */ > static int iommu_get_page_shift(u32 query_page_size) > { > - /* Supported IO page-sizes according to LoPAR */ > + /* Supported IO page-sizes according to LoPAR, note that 2M is out of order */ > const int shift[] = { > __builtin_ctzll(SZ_4K), __builtin_ctzll(SZ_64K), __builtin_ctzll(SZ_16M), > __builtin_ctzll(SZ_32M), __builtin_ctzll(SZ_64M), __builtin_ctzll(SZ_128M), > - __builtin_ctzll(SZ_256M), __builtin_ctzll(SZ_16G) > + __builtin_ctzll(SZ_256M), __builtin_ctzll(SZ_16G), __builtin_ctzll(SZ_2M) > }; > > int i = ARRAY_SIZE(shift) - 1; > + int ret = 0; > > /* > * On LoPAR, ibm,query-pe-dma-window outputs "IO Page Sizes" using a bit field: > @@ -1176,11 +1177,10 @@ static int iommu_get_page_shift(u32 query_page_size) > */ > for (; i >= 0 ; i--) { > if (query_page_size & (1 << i)) > - return shift[i]; > + ret = max(ret, shift[i]); > } > > - /* No valid page size found. */ > - return 0; > + return ret; > } > > static struct property *ddw_property_create(const char *propname, u32 liobn, u64 dma_addr, >
diff --git a/arch/powerpc/platforms/pseries/iommu.c b/arch/powerpc/platforms/pseries/iommu.c index c741689a5165..237bf405b0cb 100644 --- a/arch/powerpc/platforms/pseries/iommu.c +++ b/arch/powerpc/platforms/pseries/iommu.c @@ -1159,14 +1159,15 @@ static void reset_dma_window(struct pci_dev *dev, struct device_node *par_dn) /* Return largest page shift based on "IO Page Sizes" output of ibm,query-pe-dma-window. */ static int iommu_get_page_shift(u32 query_page_size) { - /* Supported IO page-sizes according to LoPAR */ + /* Supported IO page-sizes according to LoPAR, note that 2M is out of order */ const int shift[] = { __builtin_ctzll(SZ_4K), __builtin_ctzll(SZ_64K), __builtin_ctzll(SZ_16M), __builtin_ctzll(SZ_32M), __builtin_ctzll(SZ_64M), __builtin_ctzll(SZ_128M), - __builtin_ctzll(SZ_256M), __builtin_ctzll(SZ_16G) + __builtin_ctzll(SZ_256M), __builtin_ctzll(SZ_16G), __builtin_ctzll(SZ_2M) }; int i = ARRAY_SIZE(shift) - 1; + int ret = 0; /* * On LoPAR, ibm,query-pe-dma-window outputs "IO Page Sizes" using a bit field: @@ -1176,11 +1177,10 @@ static int iommu_get_page_shift(u32 query_page_size) */ for (; i >= 0 ; i--) { if (query_page_size & (1 << i)) - return shift[i]; + ret = max(ret, shift[i]); } - /* No valid page size found. */ - return 0; + return ret; } static struct property *ddw_property_create(const char *propname, u32 liobn, u64 dma_addr,
The upcoming PAPR spec adds a 2M page size, bit 23 right after the 16G page size in the "ibm,query-pe-dma-window" call. This adds support for the new page size. Since the new page size is out of sorted order, this changes the loop to not assume that shift[] is sorted. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> --- This might not work if PHYP keeps rejecting new window requests for less than 32768 TCEs. This is needed: https://github.com/aik/linux/commit/8cc8fa5ba5b3b4a18efbc9d81d9e5b85ca7c8a95 --- arch/powerpc/platforms/pseries/iommu.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)