The numa_fw_cfg paravirt interface is extended to include SRAT information for all hotplug-able dimms. There are 3 words for each hotplug-able memory slot, denoting start address, size and node proximity. The new info is appended after existing numa info, so that the fw_cfg layout does not break. This information is used by Seabios to build hotplug memory device objects at runtime. nb_numa_nodes is set to 1 by default (not 0), so that we always pass srat info to SeaBIOS.
v3->v4: numa_fw_cfg needs to be initalized after memory controller sets up dimm ranges. Make changes for pc_piix and pc_q35 to set numa_fw_cfg after i440fx initialization.
v2->v3: setting nb_numa_nodes to 1 is not needed
v1->v2: Dimm SRAT info (#dimms) is appended at end of existing numa fw_cfg in order not to break existing layout Documentation of the new fwcfg layout is included in docs/specs/fwcfg.txt
Signed-off-by: Vasilis Liaskovitis vasilis.liaskovitis@profitbricks.com --- docs/specs/fwcfg.txt | 28 ++++++++++++++++++++++++++++ hw/pc.c | 28 +++++++++++++++++++++++----- hw/pc.h | 1 + hw/pc_piix.c | 1 + hw/pc_q35.c | 8 +++++--- sysemu.h | 1 + 6 files changed, 59 insertions(+), 8 deletions(-) create mode 100644 docs/specs/fwcfg.txt
diff --git a/docs/specs/fwcfg.txt b/docs/specs/fwcfg.txt new file mode 100644 index 0000000..e6fcd8f --- /dev/null +++ b/docs/specs/fwcfg.txt @@ -0,0 +1,28 @@ +QEMU<->BIOS Paravirt Documentation +-------------------------------------- + +This document describes paravirt data structures passed from QEMU to BIOS. + +fw_cfg SRAT paravirt info +-------------------- +The SRAT info passed from QEMU to BIOS has the following layout: + +----------------------------------------------------------------------------------------------- +#nodes | cpu0_pxm | cpu1_pxm | ... | cpulast_pxm | node0_mem | node1_mem | ... | nodelast_mem + +----------------------------------------------------------------------------------------------- +#dimms | dimm0_start | dimm0_sz | dimm0_pxm | ... | dimmlast_start | dimmlast_sz | dimmlast_pxm + +Entry 0 contains the number of numa nodes (nb_numa_nodes). + +Entries 1..max_cpus: The next max_cpus entries describe node proximity for each +one of the vCPUs in the system. + +Entries max_cpus+1..max_cpus+nb_numa_nodes+1: The next nb_numa_nodes entries +describe the memory size for each one of the NUMA nodes in the system. + +Entry max_cpus+nb_numa_nodes+1 contains the number of memory dimms (nb_hp_dimms) + +The last 3 * nb_hp_dimms entries are organized in triplets: Each triplet contains +the physical address offset, size (in bytes), and node proximity for the +respective dimm. diff --git a/hw/pc.c b/hw/pc.c index b11e7c4..025c356 100644 --- a/hw/pc.c +++ b/hw/pc.c @@ -51,6 +51,7 @@ #include "exec-memory.h" #include "arch_init.h" #include "bitmap.h" +#include "hw/dimm.h"
/* debug PC/ISA interrupts */ //#define DEBUG_IRQ @@ -582,8 +583,6 @@ static void *bochs_bios_init(void) void *fw_cfg; uint8_t *smbios_table; size_t smbios_len; - uint64_t *numa_fw_cfg; - int i, j; PortioList *bochs_bios_port_list = g_new(PortioList, 1);
portio_list_init(bochs_bios_port_list, bochs_bios_portio_list, @@ -607,11 +606,24 @@ static void *bochs_bios_init(void)
fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, (uint8_t *)&hpet_cfg, sizeof(struct hpet_fw_config)); + + return fw_cfg; +} + +void bochs_meminfo_bios_init(void *fw_cfg) +{ + uint64_t *numa_fw_cfg; + uint64_t *hp_dimms_fw_cfg; + int i, j; + /* allocate memory for the NUMA channel: one (64bit) word for the number * of nodes, one word for each VCPU->node and one word for each node to * hold the amount of memory. + * Finally one word for the number of hotplug memory slots and three words + * for each hotplug memory slot (start address, size and node proximity). */ - numa_fw_cfg = g_malloc0((1 + max_cpus + nb_numa_nodes) * 8); + numa_fw_cfg = g_malloc0((2 + max_cpus + nb_numa_nodes + 3 * nb_hp_dimms) + * 8); numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes); for (i = 0; i < max_cpus; i++) { for (j = 0; j < nb_numa_nodes; j++) { @@ -624,10 +636,16 @@ static void *bochs_bios_init(void) for (i = 0; i < nb_numa_nodes; i++) { numa_fw_cfg[max_cpus + 1 + i] = cpu_to_le64(node_mem[i]); } + + numa_fw_cfg[1 + max_cpus + nb_numa_nodes] = cpu_to_le64(nb_hp_dimms); + + hp_dimms_fw_cfg = numa_fw_cfg + 2 + max_cpus + nb_numa_nodes; + if (nb_hp_dimms) { + dimm_setup_fwcfg_layout(hp_dimms_fw_cfg); + } fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, (uint8_t *)numa_fw_cfg, - (1 + max_cpus + nb_numa_nodes) * 8); + (2 + max_cpus + nb_numa_nodes + 3 * nb_hp_dimms) * 8);
- return fw_cfg; }
static long get_file_size(FILE *f) diff --git a/hw/pc.h b/hw/pc.h index 2237e86..075514f 100644 --- a/hw/pc.h +++ b/hw/pc.h @@ -185,5 +185,6 @@ void pc_system_firmware_init(MemoryRegion *rom_memory); #define E820_UNUSABLE 5
int e820_add_entry(uint64_t, uint64_t, uint32_t); +void bochs_meminfo_bios_init(void *fw_cfg);
#endif diff --git a/hw/pc_piix.c b/hw/pc_piix.c index fe995b9..1a99852 100644 --- a/hw/pc_piix.c +++ b/hw/pc_piix.c @@ -140,6 +140,7 @@ static void pc_init1(MemoryRegion *system_memory, i440fx_host->mch.above_4g_mem_size = above_4g_mem_size;
qdev_init_nofail(DEVICE(i440fx_host)); + bochs_meminfo_bios_init(fw_cfg); i440fx_state = &i440fx_host->mch; pci_bus = i440fx_host->parent_obj.bus; /* Xen supports additional interrupt routes from the PCI devices to diff --git a/hw/pc_q35.c b/hw/pc_q35.c index e6375bf..7ce0b53 100644 --- a/hw/pc_q35.c +++ b/hw/pc_q35.c @@ -86,6 +86,7 @@ static void pc_q35_init(QEMUMachineInitArgs *args) ICH9LPCState *ich9_lpc; PCIDevice *ahci; qemu_irq *cmos_s3; + void *fw_cfg = NULL;
pc_cpus_init(cpu_model);
@@ -111,9 +112,9 @@ static void pc_q35_init(QEMUMachineInitArgs *args)
/* allocate ram and load rom/bios */ if (!xen_enabled()) { - pc_memory_init(get_system_memory(), kernel_filename, kernel_cmdline, - initrd_filename, below_4g_mem_size, above_4g_mem_size, - rom_memory, &ram_memory); + fw_cfg = pc_memory_init(get_system_memory(), kernel_filename, + kernel_cmdline, initrd_filename, below_4g_mem_size, + above_4g_mem_size, rom_memory, &ram_memory); }
/* irq lines */ @@ -137,6 +138,7 @@ static void pc_q35_init(QEMUMachineInitArgs *args) q35_host->mch.above_4g_mem_size = above_4g_mem_size; /* pci */ qdev_init_nofail(DEVICE(q35_host)); + bochs_meminfo_bios_init(fw_cfg); host_bus = q35_host->host.pci.bus; /* create ISA bus */ lpc = pci_create_simple_multifunction(host_bus, PCI_DEVFN(ICH9_LPC_DEV, diff --git a/sysemu.h b/sysemu.h index 1b6add2..86f729e 100644 --- a/sysemu.h +++ b/sysemu.h @@ -130,6 +130,7 @@ extern QEMUClock *rtc_clock; extern int nb_numa_nodes; extern uint64_t node_mem[MAX_NODES]; extern unsigned long *node_cpumask[MAX_NODES]; +extern int nb_hp_dimms;
#define MAX_OPTION_ROMS 16 typedef struct QEMUOptionRom {