Implement __phys_to_dma() and __dma_to_phys() according to the
node id offset in 7A1000 DMA route config register.
Signed-off-by: Tiezhu Yang <[email protected]>
---
arch/mips/include/asm/mach-loongson64/boot_param.h | 1 +
arch/mips/loongson64/dma.c | 49 +++++++++++++++++++++-
arch/mips/loongson64/init.c | 13 ++++++
3 files changed, 61 insertions(+), 2 deletions(-)
diff --git a/arch/mips/include/asm/mach-loongson64/boot_param.h b/arch/mips/include/asm/mach-loongson64/boot_param.h
index 225a563..60e7a7e 100644
--- a/arch/mips/include/asm/mach-loongson64/boot_param.h
+++ b/arch/mips/include/asm/mach-loongson64/boot_param.h
@@ -218,5 +218,6 @@ struct loongson_system_configuration {
extern struct efi_memory_map_loongson *loongson_memmap;
extern struct loongson_system_configuration loongson_sysconf;
extern struct board_devices *eboard;
+extern u32 node_id_offset;
#endif
diff --git a/arch/mips/loongson64/dma.c b/arch/mips/loongson64/dma.c
index 5e86635..997c257 100644
--- a/arch/mips/loongson64/dma.c
+++ b/arch/mips/loongson64/dma.c
@@ -2,24 +2,69 @@
#include <linux/dma-direct.h>
#include <linux/init.h>
#include <linux/swiotlb.h>
+#include <boot_param.h>
-dma_addr_t __phys_to_dma(struct device *dev, phys_addr_t paddr)
+struct loongson_dma_ops {
+ dma_addr_t (*phys_to_dma)(struct device *dev, phys_addr_t paddr);
+ phys_addr_t (*dma_to_phys)(struct device *dev, dma_addr_t daddr);
+};
+
+struct loongson_dma_ops loongson_dma;
+
+dma_addr_t __rs780e_phys_to_dma(struct device *dev, phys_addr_t paddr)
{
/* We extract 2bit node id (bit 44~47, only bit 44~45 used now) from
* Loongson-3's 48bit address space and embed it into 40bit */
long nid = (paddr >> 44) & 0x3;
+
return ((nid << 44) ^ paddr) | (nid << 37);
}
-phys_addr_t __dma_to_phys(struct device *dev, dma_addr_t daddr)
+dma_addr_t __ls7a_phys_to_dma(struct device *dev, phys_addr_t paddr)
+{
+ long nid = (paddr >> 44) & 0x3;
+
+ return ((nid << 44) ^ paddr) | (nid << (36 + node_id_offset));
+}
+
+
+dma_addr_t __phys_to_dma(struct device *dev, phys_addr_t paddr)
+{
+ return loongson_dma.phys_to_dma(dev, paddr);
+}
+
+phys_addr_t __rs780e_dma_to_phys(struct device *dev, dma_addr_t daddr)
{
/* We extract 2bit node id (bit 44~47, only bit 44~45 used now) from
* Loongson-3's 48bit address space and embed it into 40bit */
long nid = (daddr >> 37) & 0x3;
+
return ((nid << 37) ^ daddr) | (nid << 44);
}
+phys_addr_t __ls7a_dma_to_phys(struct device *dev, dma_addr_t daddr)
+{
+ long nid = (daddr >> (36 + node_id_offset)) & 0x3;
+
+ return ((nid << (36 + node_id_offset)) ^ daddr) | (nid << 44);
+}
+
+phys_addr_t __dma_to_phys(struct device *dev, dma_addr_t daddr)
+{
+ return loongson_dma.dma_to_phys(dev, daddr);
+}
+
void __init plat_swiotlb_setup(void)
{
swiotlb_init(1);
+
+ if (strstr(eboard->name, "780E")) {
+ loongson_dma.phys_to_dma = __rs780e_phys_to_dma;
+ loongson_dma.dma_to_phys = __rs780e_dma_to_phys;
+ }
+
+ if (strstr(eboard->name, "7A1000")) {
+ loongson_dma.phys_to_dma = __ls7a_phys_to_dma;
+ loongson_dma.dma_to_phys = __ls7a_dma_to_phys;
+ }
}
diff --git a/arch/mips/loongson64/init.c b/arch/mips/loongson64/init.c
index 5ac1a0f..dd8463d 100644
--- a/arch/mips/loongson64/init.c
+++ b/arch/mips/loongson64/init.c
@@ -12,6 +12,11 @@
#include <asm/fw/fw.h>
#include <loongson.h>
+#include <boot_param.h>
+
+#define NODE_ID_OFFSET_ADDR 0x90000E001001041CULL
+
+u32 node_id_offset;
static void __init mips_nmi_setup(void)
{
@@ -23,6 +28,11 @@ static void __init mips_nmi_setup(void)
flush_icache_range((unsigned long)base, (unsigned long)base + 0x80);
}
+static void ls7a_early_config(void)
+{
+ node_id_offset = (*(u32 *)NODE_ID_OFFSET_ADDR >> 8) & 0x1F;
+}
+
void __init prom_init(void)
{
fw_init_cmdline();
@@ -32,6 +42,9 @@ void __init prom_init(void)
set_io_port_base((unsigned long)
ioremap(LOONGSON_PCIIO_BASE, LOONGSON_PCIIO_SIZE));
+ if (strstr(eboard->name, "7A1000"))
+ ls7a_early_config();
+
prom_init_numa_memory();
/* Hardcode to CPU UART 0 */
--
2.1.0
---- 在 星期一, 2020-03-09 16:23:24 Tiezhu Yang <[email protected]> 撰写 ----
> Implement __phys_to_dma() and __dma_to_phys() according to the
> node id offset in 7A1000 DMA route config register.
That design shocked me a lot. And It is known that some firmware didn't configure
HT Recieve window correctly to make it work. So probably for mainline kernel,
just set DMA_MASK to limit LS7A DMA address to Node0 would be a better Option?
> Signed-off-by: Tiezhu Yang <[email protected]>
> ---
> arch/mips/include/asm/mach-loongson64/boot_param.h | 1 +
> arch/mips/loongson64/dma.c | 49 +++++++++++++++++++++-
> arch/mips/loongson64/init.c | 13 ++++++
> 3 files changed, 61 insertions(+), 2 deletions(-)
>
> diff --git a/arch/mips/include/asm/mach-loongson64/boot_param.h b/arch/mips/include/asm/mach-loongson64/boot_param.h
> index 225a563..60e7a7e 100644
> --- a/arch/mips/include/asm/mach-loongson64/boot_param.h
> +++ b/arch/mips/include/asm/mach-loongson64/boot_param.h
> @@ -218,5 +218,6 @@ struct loongson_system_configuration {
> extern struct efi_memory_map_loongson *loongson_memmap;
> extern struct loongson_system_configuration loongson_sysconf;
> extern struct board_devices *eboard;
> +extern u32 node_id_offset;
>
> #endif
> diff --git a/arch/mips/loongson64/dma.c b/arch/mips/loongson64/dma.c
> index 5e86635..997c257 100644
> --- a/arch/mips/loongson64/dma.c
> +++ b/arch/mips/loongson64/dma.c
> @@ -2,24 +2,69 @@
> #include <linux/dma-direct.h>
> #include <linux/init.h>
> #include <linux/swiotlb.h>
> +#include <boot_param.h>
>
> -dma_addr_t __phys_to_dma(struct device *dev, phys_addr_t paddr)
> +struct loongson_dma_ops {
> + dma_addr_t (*phys_to_dma)(struct device *dev, phys_addr_t paddr);
> + phys_addr_t (*dma_to_phys)(struct device *dev, dma_addr_t daddr);
> +};
> +
> +struct loongson_dma_ops loongson_dma;
> +
> +dma_addr_t __rs780e_phys_to_dma(struct device *dev, phys_addr_t paddr)
> {
> /* We extract 2bit node id (bit 44~47, only bit 44~45 used now) from
> * Loongson-3's 48bit address space and embed it into 40bit */
> long nid = (paddr >> 44) & 0x3;
> +
> return ((nid << 44) ^ paddr) | (nid << 37);
> }
>
> -phys_addr_t __dma_to_phys(struct device *dev, dma_addr_t daddr)
> +dma_addr_t __ls7a_phys_to_dma(struct device *dev, phys_addr_t paddr)
> +{
> + long nid = (paddr >> 44) & 0x3;
> +
> + return ((nid << 44) ^ paddr) | (nid << (36 + node_id_offset));
> +}
> +
> +
> +dma_addr_t __phys_to_dma(struct device *dev, phys_addr_t paddr)
> +{
> + return loongson_dma.phys_to_dma(dev, paddr);
> +}
> +
> +phys_addr_t __rs780e_dma_to_phys(struct device *dev, dma_addr_t daddr)
> {
> /* We extract 2bit node id (bit 44~47, only bit 44~45 used now) from
> * Loongson-3's 48bit address space and embed it into 40bit */
> long nid = (daddr >> 37) & 0x3;
> +
> return ((nid << 37) ^ daddr) | (nid << 44);
> }
>
> +phys_addr_t __ls7a_dma_to_phys(struct device *dev, dma_addr_t daddr)
> +{
> + long nid = (daddr >> (36 + node_id_offset)) & 0x3;
> +
> + return ((nid << (36 + node_id_offset)) ^ daddr) | (nid << 44);
> +}
> +
> +phys_addr_t __dma_to_phys(struct device *dev, dma_addr_t daddr)
> +{
> + return loongson_dma.dma_to_phys(dev, daddr);
> +}
> +
> void __init plat_swiotlb_setup(void)
> {
> swiotlb_init(1);
> +
> + if (strstr(eboard->name, "780E")) {
> + loongson_dma.phys_to_dma = __rs780e_phys_to_dma;
> + loongson_dma.dma_to_phys = __rs780e_dma_to_phys;
> + }
> +
> + if (strstr(eboard->name, "7A1000")) {
> + loongson_dma.phys_to_dma = __ls7a_phys_to_dma;
> + loongson_dma.dma_to_phys = __ls7a_dma_to_phys;
> + }
> }
> diff --git a/arch/mips/loongson64/init.c b/arch/mips/loongson64/init.c
> index 5ac1a0f..dd8463d 100644
> --- a/arch/mips/loongson64/init.c
> +++ b/arch/mips/loongson64/init.c
> @@ -12,6 +12,11 @@
> #include <asm/fw/fw.h>
>
> #include <loongson.h>
> +#include <boot_param.h>
> +
> +#define NODE_ID_OFFSET_ADDR 0x90000E001001041CULL
> +
> +u32 node_id_offset;
>
> static void __init mips_nmi_setup(void)
> {
> @@ -23,6 +28,11 @@ static void __init mips_nmi_setup(void)
> flush_icache_range((unsigned long)base, (unsigned long)base + 0x80);
> }
>
> +static void ls7a_early_config(void)
> +{
> + node_id_offset = (*(u32 *)NODE_ID_OFFSET_ADDR >> 8) & 0x1F;
Please avoid raw pointer. Use readl/writel instead.
> +}
> +
> void __init prom_init(void)
> {
> fw_init_cmdline();
> @@ -32,6 +42,9 @@ void __init prom_init(void)
> set_io_port_base((unsigned long)
> ioremap(LOONGSON_PCIIO_BASE, LOONGSON_PCIIO_SIZE));
>
> + if (strstr(eboard->name, "7A1000"))
> + ls7a_early_config();
> +
> prom_init_numa_memory();
>
> /* Hardcode to CPU UART 0 */
> --
> 2.1.0
>
>
On 03/09/2020 04:44 PM, Jiaxun Yang wrote:
>
> ---- 在 星期一, 2020-03-09 16:23:24 Tiezhu Yang <[email protected]> 撰写 ----
> > Implement __phys_to_dma() and __dma_to_phys() according to the
> > node id offset in 7A1000 DMA route config register.
>
> That design shocked me a lot. And It is known that some firmware didn't configure
> HT Recieve window correctly to make it work. So probably for mainline kernel,
> just set DMA_MASK to limit LS7A DMA address to Node0 would be a better Option?
Hi Jiaxun,
Let me rethink it and find a proper way.
>
> > Signed-off-by: Tiezhu Yang <[email protected]>
> > ---
> > arch/mips/include/asm/mach-loongson64/boot_param.h | 1 +
> > arch/mips/loongson64/dma.c | 49 +++++++++++++++++++++-
> > arch/mips/loongson64/init.c | 13 ++++++
> > 3 files changed, 61 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/mips/include/asm/mach-loongson64/boot_param.h b/arch/mips/include/asm/mach-loongson64/boot_param.h
> > index 225a563..60e7a7e 100644
> > --- a/arch/mips/include/asm/mach-loongson64/boot_param.h
> > +++ b/arch/mips/include/asm/mach-loongson64/boot_param.h
> > @@ -218,5 +218,6 @@ struct loongson_system_configuration {
> > extern struct efi_memory_map_loongson *loongson_memmap;
> > extern struct loongson_system_configuration loongson_sysconf;
> > extern struct board_devices *eboard;
> > +extern u32 node_id_offset;
> >
> > #endif
> > diff --git a/arch/mips/loongson64/dma.c b/arch/mips/loongson64/dma.c
> > index 5e86635..997c257 100644
> > --- a/arch/mips/loongson64/dma.c
> > +++ b/arch/mips/loongson64/dma.c
> > @@ -2,24 +2,69 @@
> > #include <linux/dma-direct.h>
> > #include <linux/init.h>
> > #include <linux/swiotlb.h>
> > +#include <boot_param.h>
> >
> > -dma_addr_t __phys_to_dma(struct device *dev, phys_addr_t paddr)
> > +struct loongson_dma_ops {
> > + dma_addr_t (*phys_to_dma)(struct device *dev, phys_addr_t paddr);
> > + phys_addr_t (*dma_to_phys)(struct device *dev, dma_addr_t daddr);
> > +};
> > +
> > +struct loongson_dma_ops loongson_dma;
> > +
> > +dma_addr_t __rs780e_phys_to_dma(struct device *dev, phys_addr_t paddr)
> > {
> > /* We extract 2bit node id (bit 44~47, only bit 44~45 used now) from
> > * Loongson-3's 48bit address space and embed it into 40bit */
> > long nid = (paddr >> 44) & 0x3;
> > +
> > return ((nid << 44) ^ paddr) | (nid << 37);
> > }
> >
> > -phys_addr_t __dma_to_phys(struct device *dev, dma_addr_t daddr)
> > +dma_addr_t __ls7a_phys_to_dma(struct device *dev, phys_addr_t paddr)
> > +{
> > + long nid = (paddr >> 44) & 0x3;
> > +
> > + return ((nid << 44) ^ paddr) | (nid << (36 + node_id_offset));
> > +}
> > +
> > +
> > +dma_addr_t __phys_to_dma(struct device *dev, phys_addr_t paddr)
> > +{
> > + return loongson_dma.phys_to_dma(dev, paddr);
> > +}
> > +
> > +phys_addr_t __rs780e_dma_to_phys(struct device *dev, dma_addr_t daddr)
> > {
> > /* We extract 2bit node id (bit 44~47, only bit 44~45 used now) from
> > * Loongson-3's 48bit address space and embed it into 40bit */
> > long nid = (daddr >> 37) & 0x3;
> > +
> > return ((nid << 37) ^ daddr) | (nid << 44);
> > }
> >
> > +phys_addr_t __ls7a_dma_to_phys(struct device *dev, dma_addr_t daddr)
> > +{
> > + long nid = (daddr >> (36 + node_id_offset)) & 0x3;
> > +
> > + return ((nid << (36 + node_id_offset)) ^ daddr) | (nid << 44);
> > +}
> > +
> > +phys_addr_t __dma_to_phys(struct device *dev, dma_addr_t daddr)
> > +{
> > + return loongson_dma.dma_to_phys(dev, daddr);
> > +}
> > +
> > void __init plat_swiotlb_setup(void)
> > {
> > swiotlb_init(1);
> > +
> > + if (strstr(eboard->name, "780E")) {
> > + loongson_dma.phys_to_dma = __rs780e_phys_to_dma;
> > + loongson_dma.dma_to_phys = __rs780e_dma_to_phys;
> > + }
> > +
> > + if (strstr(eboard->name, "7A1000")) {
> > + loongson_dma.phys_to_dma = __ls7a_phys_to_dma;
> > + loongson_dma.dma_to_phys = __ls7a_dma_to_phys;
> > + }
> > }
> > diff --git a/arch/mips/loongson64/init.c b/arch/mips/loongson64/init.c
> > index 5ac1a0f..dd8463d 100644
> > --- a/arch/mips/loongson64/init.c
> > +++ b/arch/mips/loongson64/init.c
> > @@ -12,6 +12,11 @@
> > #include <asm/fw/fw.h>
> >
> > #include <loongson.h>
> > +#include <boot_param.h>
> > +
> > +#define NODE_ID_OFFSET_ADDR 0x90000E001001041CULL
> > +
> > +u32 node_id_offset;
> >
> > static void __init mips_nmi_setup(void)
> > {
> > @@ -23,6 +28,11 @@ static void __init mips_nmi_setup(void)
> > flush_icache_range((unsigned long)base, (unsigned long)base + 0x80);
> > }
> >
> > +static void ls7a_early_config(void)
> > +{
> > + node_id_offset = (*(u32 *)NODE_ID_OFFSET_ADDR >> 8) & 0x1F;
>
> Please avoid raw pointer. Use readl/writel instead.
OK, I will do it.
Thanks,
Tiezhu Yang
>
> > +}
> > +
> > void __init prom_init(void)
> > {
> > fw_init_cmdline();
> > @@ -32,6 +42,9 @@ void __init prom_init(void)
> > set_io_port_base((unsigned long)
> > ioremap(LOONGSON_PCIIO_BASE, LOONGSON_PCIIO_SIZE));
> >
> > + if (strstr(eboard->name, "7A1000"))
> > + ls7a_early_config();
> > +
> > prom_init_numa_memory();
> >
> > /* Hardcode to CPU UART 0 */
> > --
> > 2.1.0
> >
> >
On Mon, Mar 09, 2020 at 04:23:24PM +0800, Tiezhu Yang wrote:
> Implement __phys_to_dma() and __dma_to_phys() according to the
> node id offset in 7A1000 DMA route config register.
Can you just switch Loongson over to use the dma_pfn_offset field in
struct device? I'd love to kill the __phys_to_dma and __dma_to_phys
hooks wherever possible.
On 03/13/2020 04:24 PM, Christoph Hellwig wrote:
> On Mon, Mar 09, 2020 at 04:23:24PM +0800, Tiezhu Yang wrote:
>> Implement __phys_to_dma() and __dma_to_phys() according to the
>> node id offset in 7A1000 DMA route config register.
> Can you just switch Loongson over to use the dma_pfn_offset field in
> struct device? I'd love to kill the __phys_to_dma and __dma_to_phys
> hooks wherever possible.
Hi Christoph,
Thanks for your suggestion, I will refactor the code.
Thanks,
Tiezhu Yang