On 10/28/23 06:49, Andrew Jones wrote:
On Sat, Oct 28, 2023 at 05:54:16AM -0300, Daniel Henrique Barboza wrote:QEMU already implements zicbom (Cache Block Management Operations) and zicboz (Cache Block Zero Operations). Commit 59cb29d6a5 ("target/riscv: add Zicbop cbo.prefetch{i, r, m} placeholder") added placeholders for what would be the instructions for zicbop (Cache Block Prefetch Operations), which are now no-ops. The RVA22U64 profile mandates zicbop, which means that applications that run with this profile might expect zicbop to be present in the riscv,isa DT and might behave badly if it's absent. Adding zicbop as an extension will make our future RVA22U64 implementation more in line with what userspace expects and, if/when cache block prefetch operations became relevant to QEMU, we already have the extension flag to turn then on/off as needed. Signed-off-by: Daniel Henrique Barboza <dbarb...@ventanamicro.com> --- hw/riscv/virt.c | 5 +++++ target/riscv/cpu.c | 3 +++ target/riscv/cpu_cfg.h | 2 ++ 3 files changed, 10 insertions(+) diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c index 1732c42915..99c087240f 100644 --- a/hw/riscv/virt.c +++ b/hw/riscv/virt.c @@ -273,6 +273,11 @@ static void create_fdt_socket_cpus(RISCVVirtState *s, int socket, cpu_ptr->cfg.cboz_blocksize); }+ if (cpu_ptr->cfg.ext_zicbop) {+ qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cbop-block-size",I think we need to get this node approved by devicet...@vger.kernel.org and merged into [1] first. [1] Linux repo: Documentation/devicetree/bindings/riscv/cpus.yaml
Ouch ... I wasn't expecting that :( We can't add zicbop without the blocksize DT and then claim that we're implement rva22u64. It'll be an 'almost implementation' of the profile because Zicbop is 'almost supported'. I'll send a DT patch for cbop-block-size and see how that goes. We would need an ack for the DT folks that we're not adding the wrong attribute for cbop. If we can't have an ack in time we should just postpone profile support for next release. I'll send a v7 to keep reviews rolling on our side but now it seems like we have a kernel dependency to merge this. Thanks, Daniel
+ cpu_ptr->cfg.cbop_blocksize); + } + qemu_fdt_setprop_string(ms->fdt, cpu_name, "compatible", "riscv"); qemu_fdt_setprop_string(ms->fdt, cpu_name, "status", "okay"); qemu_fdt_setprop_cell(ms->fdt, cpu_name, "reg", diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c index f40da4c661..6c0050988f 100644 --- a/target/riscv/cpu.c +++ b/target/riscv/cpu.c @@ -78,6 +78,7 @@ const uint32_t misa_bits[] = {RVI, RVE, RVM, RVA, RVF, RVD, RVV, */ const RISCVIsaExtData isa_edata_arr[] = { ISA_EXT_DATA_ENTRY(zicbom, PRIV_VERSION_1_12_0, ext_zicbom), + ISA_EXT_DATA_ENTRY(zicbop, PRIV_VERSION_1_12_0, ext_zicbop), ISA_EXT_DATA_ENTRY(zicboz, PRIV_VERSION_1_12_0, ext_zicboz), ISA_EXT_DATA_ENTRY(zicond, PRIV_VERSION_1_12_0, ext_zicond), ISA_EXT_DATA_ENTRY(zicntr, PRIV_VERSION_1_12_0, ext_zicntr), @@ -1336,6 +1337,7 @@ const RISCVCPUMultiExtConfig riscv_cpu_extensions[] = { MULTI_EXT_CFG_BOOL("zhinxmin", ext_zhinxmin, false),MULTI_EXT_CFG_BOOL("zicbom", ext_zicbom, true),+ MULTI_EXT_CFG_BOOL("zicbop", ext_zicbop, true), MULTI_EXT_CFG_BOOL("zicboz", ext_zicboz, true),MULTI_EXT_CFG_BOOL("zmmul", ext_zmmul, false),@@ -1424,6 +1426,7 @@ Property riscv_cpu_options[] = { DEFINE_PROP_UINT16("elen", RISCVCPU, cfg.elen, 64),DEFINE_PROP_UINT16("cbom_blocksize", RISCVCPU, cfg.cbom_blocksize, 64),+ DEFINE_PROP_UINT16("cbop_blocksize", RISCVCPU, cfg.cbop_blocksize, 64), DEFINE_PROP_UINT16("cboz_blocksize", RISCVCPU, cfg.cboz_blocksize, 64),DEFINE_PROP_END_OF_LIST(),diff --git a/target/riscv/cpu_cfg.h b/target/riscv/cpu_cfg.h index 6eef4a51ea..2203b4c45b 100644 --- a/target/riscv/cpu_cfg.h +++ b/target/riscv/cpu_cfg.h @@ -65,6 +65,7 @@ struct RISCVCPUConfig { bool ext_zicntr; bool ext_zicsr; bool ext_zicbom; + bool ext_zicbop; bool ext_zicboz; bool ext_zicond; bool ext_zihintntl; @@ -134,6 +135,7 @@ struct RISCVCPUConfig { uint16_t vlen; uint16_t elen; uint16_t cbom_blocksize; + uint16_t cbop_blocksize; uint16_t cboz_blocksize; bool mmu; bool pmp; -- 2.41.0Otherwise, Reviewed-by: Andrew Jones <ajo...@ventanamicro.com> Thanks, drew
