| Message ID | 20210923120236.3692135-33-wei.chen@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Add device tree based NUMA support to Arm | expand |
On Thu, 23 Sep 2021, Wei Chen wrote: > In this API, we scan whole device tree to parse CPU node id, memory ^ function ^ the whole > node id and distance-map. Though early_scan_node will invoke has a > handler to process memory nodes. If we want to parse memory node id > in this handler, we have to embeded NUMA parse code in this handler. ^ embed > But we still need to scan whole device tree to find CPU NUMA id and > distance-map. In this case, we include memory NUMA id parse in this > API too. Another benefit is that we have a unique entry for device ^ function > tree NUMA data parse. Ah, that's the explanation I was asking for earlier! > Signed-off-by: Wei Chen <wei.chen@arm.com> > --- > xen/arch/arm/numa_device_tree.c | 30 ++++++++++++++++++++++++++++++ > xen/include/asm-arm/numa.h | 1 + > 2 files changed, 31 insertions(+) > > diff --git a/xen/arch/arm/numa_device_tree.c b/xen/arch/arm/numa_device_tree.c > index e7fa84df4c..6a3fed0002 100644 > --- a/xen/arch/arm/numa_device_tree.c > +++ b/xen/arch/arm/numa_device_tree.c > @@ -242,3 +242,33 @@ static int __init fdt_parse_numa_distance_map_v1(const void *fdt, int node) > > return 0; > } > + > +static int __init fdt_scan_numa_nodes(const void *fdt, > + int node, const char *uname, int depth, > + u32 address_cells, u32 size_cells, void *data) Please align parameters > +{ > + int len, ret = 0; > + const void *prop; > + > + prop = fdt_getprop(fdt, node, "device_type", &len); > + if (prop) code style > + { > + len += 1; > + if ( memcmp(prop, "cpu", len) == 0 ) > + ret = fdt_parse_numa_cpu_node(fdt, node); > + else if ( memcmp(prop, "memory", len) == 0 ) > + ret = fdt_parse_numa_memory_node(fdt, node, uname, > + address_cells, size_cells); I realize that with the inclusion of '\0' in the check, the usage of memcmp should be safe, but I would prefer if we used strncmp instead. > + } > + else if ( fdt_node_check_compatible(fdt, node, > + "numa-distance-map-v1") == 0 ) > + ret = fdt_parse_numa_distance_map_v1(fdt, node); > + > + return ret; > +} > + > +/* Initialize NUMA from device tree */ > +int __init numa_device_tree_init(const void *fdt) > +{ > + return device_tree_for_each_node(fdt, 0, fdt_scan_numa_nodes, NULL); > +} > diff --git a/xen/include/asm-arm/numa.h b/xen/include/asm-arm/numa.h > index 7675012cb7..f46e8e2935 100644 > --- a/xen/include/asm-arm/numa.h > +++ b/xen/include/asm-arm/numa.h > @@ -23,6 +23,7 @@ typedef u8 nodeid_t; > #define NR_NODE_MEMBLKS NR_MEM_BANKS > > extern void numa_set_distance(nodeid_t from, nodeid_t to, uint32_t distance); > +extern int numa_device_tree_init(const void *fdt); > > #else > > -- > 2.25.1 >
Hi Stefano, > -----Original Message----- > From: Stefano Stabellini <sstabellini@kernel.org> > Sent: 2021年9月24日 11:17 > To: Wei Chen <Wei.Chen@arm.com> > Cc: xen-devel@lists.xenproject.org; sstabellini@kernel.org; julien@xen.org; > Bertrand Marquis <Bertrand.Marquis@arm.com> > Subject: Re: [PATCH 32/37] xen/arm: unified entry to parse all NUMA data > from device tree > > On Thu, 23 Sep 2021, Wei Chen wrote: > > In this API, we scan whole device tree to parse CPU node id, memory > ^ function ^ the whole > > > node id and distance-map. Though early_scan_node will invoke has a > > handler to process memory nodes. If we want to parse memory node id > > in this handler, we have to embeded NUMA parse code in this handler. > ^ embed > > > But we still need to scan whole device tree to find CPU NUMA id and > > distance-map. In this case, we include memory NUMA id parse in this > > API too. Another benefit is that we have a unique entry for device > ^ function > > > tree NUMA data parse. > > Ah, that's the explanation I was asking for earlier! > The question about device_tree_get_meminfo? > > > Signed-off-by: Wei Chen <wei.chen@arm.com> > > --- > > xen/arch/arm/numa_device_tree.c | 30 ++++++++++++++++++++++++++++++ > > xen/include/asm-arm/numa.h | 1 + > > 2 files changed, 31 insertions(+) > > > > diff --git a/xen/arch/arm/numa_device_tree.c > b/xen/arch/arm/numa_device_tree.c > > index e7fa84df4c..6a3fed0002 100644 > > --- a/xen/arch/arm/numa_device_tree.c > > +++ b/xen/arch/arm/numa_device_tree.c > > @@ -242,3 +242,33 @@ static int __init > fdt_parse_numa_distance_map_v1(const void *fdt, int node) > > > > return 0; > > } > > + > > +static int __init fdt_scan_numa_nodes(const void *fdt, > > + int node, const char *uname, int depth, > > + u32 address_cells, u32 size_cells, void *data) > > Please align parameters > OK > > > +{ > > + int len, ret = 0; > > + const void *prop; > > + > > + prop = fdt_getprop(fdt, node, "device_type", &len); > > + if (prop) > > code style > OK > > > + { > > + len += 1; > > + if ( memcmp(prop, "cpu", len) == 0 ) > > + ret = fdt_parse_numa_cpu_node(fdt, node); > > + else if ( memcmp(prop, "memory", len) == 0 ) > > + ret = fdt_parse_numa_memory_node(fdt, node, uname, > > + address_cells, size_cells); > > I realize that with the inclusion of '\0' in the check, the usage of > memcmp should be safe, but I would prefer if we used strncmp instead. > Ok, I will use strncmp in next version. > > > + } > > + else if ( fdt_node_check_compatible(fdt, node, > > + "numa-distance-map-v1") == 0 ) > > + ret = fdt_parse_numa_distance_map_v1(fdt, node); > > + > > + return ret; > > +} > > + > > +/* Initialize NUMA from device tree */ > > +int __init numa_device_tree_init(const void *fdt) > > +{ > > + return device_tree_for_each_node(fdt, 0, fdt_scan_numa_nodes, NULL); > > +} > > diff --git a/xen/include/asm-arm/numa.h b/xen/include/asm-arm/numa.h > > index 7675012cb7..f46e8e2935 100644 > > --- a/xen/include/asm-arm/numa.h > > +++ b/xen/include/asm-arm/numa.h > > @@ -23,6 +23,7 @@ typedef u8 nodeid_t; > > #define NR_NODE_MEMBLKS NR_MEM_BANKS > > > > extern void numa_set_distance(nodeid_t from, nodeid_t to, uint32_t > distance); > > +extern int numa_device_tree_init(const void *fdt); > > > > #else > > > > -- > > 2.25.1 > >
On Fri, 24 Sep 2021, Wei Chen wrote: > > -----Original Message----- > > From: Stefano Stabellini <sstabellini@kernel.org> > > Sent: 2021年9月24日 11:17 > > To: Wei Chen <Wei.Chen@arm.com> > > Cc: xen-devel@lists.xenproject.org; sstabellini@kernel.org; julien@xen.org; > > Bertrand Marquis <Bertrand.Marquis@arm.com> > > Subject: Re: [PATCH 32/37] xen/arm: unified entry to parse all NUMA data > > from device tree > > > > On Thu, 23 Sep 2021, Wei Chen wrote: > > > In this API, we scan whole device tree to parse CPU node id, memory > > ^ function ^ the whole > > > > > node id and distance-map. Though early_scan_node will invoke has a > > > handler to process memory nodes. If we want to parse memory node id > > > in this handler, we have to embeded NUMA parse code in this handler. > > ^ embed > > > > > But we still need to scan whole device tree to find CPU NUMA id and > > > distance-map. In this case, we include memory NUMA id parse in this > > > API too. Another benefit is that we have a unique entry for device > > ^ function > > > > > tree NUMA data parse. > > > > Ah, that's the explanation I was asking for earlier! > > > > The question about device_tree_get_meminfo? Yes, it would be nice to reuse process_memory_node if we can, but I understand if we cannot.
diff --git a/xen/arch/arm/numa_device_tree.c b/xen/arch/arm/numa_device_tree.c index e7fa84df4c..6a3fed0002 100644 --- a/xen/arch/arm/numa_device_tree.c +++ b/xen/arch/arm/numa_device_tree.c @@ -242,3 +242,33 @@ static int __init fdt_parse_numa_distance_map_v1(const void *fdt, int node) return 0; } + +static int __init fdt_scan_numa_nodes(const void *fdt, + int node, const char *uname, int depth, + u32 address_cells, u32 size_cells, void *data) +{ + int len, ret = 0; + const void *prop; + + prop = fdt_getprop(fdt, node, "device_type", &len); + if (prop) + { + len += 1; + if ( memcmp(prop, "cpu", len) == 0 ) + ret = fdt_parse_numa_cpu_node(fdt, node); + else if ( memcmp(prop, "memory", len) == 0 ) + ret = fdt_parse_numa_memory_node(fdt, node, uname, + address_cells, size_cells); + } + else if ( fdt_node_check_compatible(fdt, node, + "numa-distance-map-v1") == 0 ) + ret = fdt_parse_numa_distance_map_v1(fdt, node); + + return ret; +} + +/* Initialize NUMA from device tree */ +int __init numa_device_tree_init(const void *fdt) +{ + return device_tree_for_each_node(fdt, 0, fdt_scan_numa_nodes, NULL); +} diff --git a/xen/include/asm-arm/numa.h b/xen/include/asm-arm/numa.h index 7675012cb7..f46e8e2935 100644 --- a/xen/include/asm-arm/numa.h +++ b/xen/include/asm-arm/numa.h @@ -23,6 +23,7 @@ typedef u8 nodeid_t; #define NR_NODE_MEMBLKS NR_MEM_BANKS extern void numa_set_distance(nodeid_t from, nodeid_t to, uint32_t distance); +extern int numa_device_tree_init(const void *fdt); #else
In this API, we scan whole device tree to parse CPU node id, memory node id and distance-map. Though early_scan_node will invoke has a handler to process memory nodes. If we want to parse memory node id in this handler, we have to embeded NUMA parse code in this handler. But we still need to scan whole device tree to find CPU NUMA id and distance-map. In this case, we include memory NUMA id parse in this API too. Another benefit is that we have a unique entry for device tree NUMA data parse. Signed-off-by: Wei Chen <wei.chen@arm.com> --- xen/arch/arm/numa_device_tree.c | 30 ++++++++++++++++++++++++++++++ xen/include/asm-arm/numa.h | 1 + 2 files changed, 31 insertions(+)