| Message ID | 20160114095933.13778.32364.sendpatchset@little-apple (mailing list archive) |
|---|---|
| State | Under Review |
| Delegated to: | Geert Uytterhoeven |
| Headers | show |
Hi Magnus, Thank you for the patch. On Thursday 14 January 2016 18:59:33 Magnus Damm wrote: > From: Magnus Damm <damm+renesas@opensource.se> > > Update the rcar-dmac driver to rework comments and variable names > related to lowmem and highmem. > > Terminology such as highmem and lowmem are commonly used for Linux > kernel memory managment. Highmem in particular is about handling more > physical memory than can fit into the kernel virtual address space. > > Highmem and lowmem are related to but not equivalent to the 40-bit > address limitation that is the real issue the rcar-dmac driver has to > deal with. For instance, with 1 GiB of RAM highmem is usually needed > on a 32-bit architecture, but a 32-bit address is still enough. > > 64-bit architectures tend to not use highmem since the virtual > address space is more than enough for loads of lowmem. 32-bit > architectures may still need highmem to access all memory. > > The rcar-dmac driver and the DMA device needs to handle the 40-bit > limitation on both 64-bit and 32-bit architectures. > > This patch renames the highmem variable to fourty_bit_required to > avoid confusion. And the comment gets updated as well. > > Signed-off-by: Magnus Damm <damm+renesas@opensource.se> > --- > > drivers/dma/sh/rcar-dmac.c | 12 ++++++------ > 1 file changed, 6 insertions(+), 6 deletions(-) > > --- 0001/drivers/dma/sh/rcar-dmac.c > +++ work/drivers/dma/sh/rcar-dmac.c 2016-01-14 18:35:23.780513000 +0900 > @@ -889,7 +889,7 @@ rcar_dmac_chan_prep_sg(struct rcar_dmac_ > unsigned int nchunks = 0; > unsigned int max_chunk_size; > unsigned int full_size = 0; > - bool highmem = false; > + bool fourty_bit_required = false; s/fourty_bit_required/fourty_bits_required/ ? Or, as the variable name seems a bit long, need_40bits ? > unsigned int i; > > desc = rcar_dmac_desc_get(chan); > @@ -935,7 +935,7 @@ rcar_dmac_chan_prep_sg(struct rcar_dmac_ > * hardware descriptor lists. > */ > if (dev_addr >> 32 || mem_addr >> 32) > - highmem = true; > + fourty_bit_required = true; > #endif > > chunk = rcar_dmac_xfer_chunk_get(chan); > @@ -977,13 +977,13 @@ rcar_dmac_chan_prep_sg(struct rcar_dmac_ > * Use hardware descriptor lists if possible when more than one chunk > * needs to be transferred (otherwise they don't make much sense). > * > - * The highmem check currently covers the whole transfer. As an > - * optimization we could use descriptor lists for consecutive lowmem > - * chunks and direct manual mode for highmem chunks. Whether the > + * The 40-bit address check currently covers the whole transfer. As an > + * optimization we could use descriptor lists for consecutive 32-bit > + * chunks and direct manual mode for 40-bit chunks. Whether the > * performance improvement would be significant enough compared to the > * additional complexity remains to be investigated. > */ > - desc->hwdescs.use = !highmem && nchunks > 1; > + desc->hwdescs.use = !fourty_bit_required && nchunks > 1; > if (desc->hwdescs.use) { > if (rcar_dmac_fill_hwdesc(chan, desc) < 0) > desc->hwdescs.use = false;
--- 0001/drivers/dma/sh/rcar-dmac.c +++ work/drivers/dma/sh/rcar-dmac.c 2016-01-14 18:35:23.780513000 +0900 @@ -889,7 +889,7 @@ rcar_dmac_chan_prep_sg(struct rcar_dmac_ unsigned int nchunks = 0; unsigned int max_chunk_size; unsigned int full_size = 0; - bool highmem = false; + bool fourty_bit_required = false; unsigned int i; desc = rcar_dmac_desc_get(chan); @@ -935,7 +935,7 @@ rcar_dmac_chan_prep_sg(struct rcar_dmac_ * hardware descriptor lists. */ if (dev_addr >> 32 || mem_addr >> 32) - highmem = true; + fourty_bit_required = true; #endif chunk = rcar_dmac_xfer_chunk_get(chan); @@ -977,13 +977,13 @@ rcar_dmac_chan_prep_sg(struct rcar_dmac_ * Use hardware descriptor lists if possible when more than one chunk * needs to be transferred (otherwise they don't make much sense). * - * The highmem check currently covers the whole transfer. As an - * optimization we could use descriptor lists for consecutive lowmem - * chunks and direct manual mode for highmem chunks. Whether the + * The 40-bit address check currently covers the whole transfer. As an + * optimization we could use descriptor lists for consecutive 32-bit + * chunks and direct manual mode for 40-bit chunks. Whether the * performance improvement would be significant enough compared to the * additional complexity remains to be investigated. */ - desc->hwdescs.use = !highmem && nchunks > 1; + desc->hwdescs.use = !fourty_bit_required && nchunks > 1; if (desc->hwdescs.use) { if (rcar_dmac_fill_hwdesc(chan, desc) < 0) desc->hwdescs.use = false;