1 Compute Express Link (CXL) 2 ========================== 3 From the view of a single host, CXL is an interconnect standard that 4 targets accelerators and memory devices attached to a CXL host. 5 This description will focus on those aspects visible either to 6 software running on a QEMU emulated host or to the internals of 7 functional emulation. As such, it will skip over many of the 8 electrical and protocol elements that would be more of interest 9 for real hardware and will dominate more general introductions to CXL. 10 It will also completely ignore the fabric management aspects of CXL 11 by considering only a single host and a static configuration. 12 13 CXL shares many concepts and much of the infrastructure of PCI Express, 14 with CXL Host Bridges, which have CXL Root Ports which may be directly 15 attached to CXL or PCI End Points. Alternatively there may be CXL Switches 16 with CXL and PCI Endpoints attached below them. In many cases additional 17 control and capabilities are exposed via PCI Express interfaces. 18 This sharing of interfaces and hence emulation code is reflected 19 in how the devices are emulated in QEMU. In most cases the various 20 CXL elements are built upon an equivalent PCIe devices. 21 22 CXL devices support the following interfaces: 23 24 * Most conventional PCIe interfaces 25 26 - Configuration space access 27 - BAR mapped memory accesses used for registers and mailboxes. 28 - MSI/MSI-X 29 - AER 30 - DOE mailboxes 31 - IDE 32 - Many other PCI express defined interfaces.. 33 34 * Memory operations 35 36 - Equivalent of accessing DRAM / NVDIMMs. Any access / feature 37 supported by the host for normal memory should also work for 38 CXL attached memory devices. 39 40 * Cache operations. The are mostly irrelevant to QEMU emulation as 41 QEMU is not emulating a coherency protocol. Any emulation related 42 to these will be device specific and is out of the scope of this 43 document. 44 45 CXL 2.0 Device Types 46 -------------------- 47 CXL 2.0 End Points are often categorized into three types. 48 49 **Type 1:** These support coherent caching of host memory. Example might 50 be a crypto accelerators. May also have device private memory accessible 51 via means such as PCI memory reads and writes to BARs. 52 53 **Type 2:** These support coherent caching of host memory and host 54 managed device memory (HDM) for which the coherency protocol is managed 55 by the host. This is a complex topic, so for more information on CXL 56 coherency see the CXL 2.0 specification. 57 58 **Type 3 Memory devices:** These devices act as a means of attaching 59 additional memory (HDM) to a CXL host including both volatile and 60 persistent memory. The CXL topology may support interleaving across a 61 number of Type 3 memory devices using HDM Decoders in the host, host 62 bridge, switch upstream port and endpoints. 63 64 Scope of CXL emulation in QEMU 65 ------------------------------ 66 The focus of CXL emulation is CXL revision 2.0 and later. Earlier CXL 67 revisions defined a smaller set of features, leaving much of the control 68 interface as implementation defined or device specific, making generic 69 emulation challenging with host specific firmware being responsible 70 for setup and the Endpoints being presented to operating systems 71 as Root Complex Integrated End Points. CXL rev 2.0 looks a lot 72 more like PCI Express, with fully specified discoverability 73 of the CXL topology. 74 75 CXL System components 76 ---------------------- 77 A CXL system is made up a Host with a number of 'standard components' 78 the control and capabilities of which are discoverable by system software 79 using means described in the CXL 2.0 specification. 80 81 CXL Fixed Memory Windows (CFMW) 82 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 83 A CFMW consists of a particular range of Host Physical Address space 84 which is routed to particular CXL Host Bridges. At time of generic 85 software initialization it will have a particularly interleaving 86 configuration and associated Quality of Service Throttling Group (QTG). 87 This information is available to system software, when making 88 decisions about how to configure interleave across available CXL 89 memory devices. It is provide as CFMW Structures (CFMWS) in 90 the CXL Early Discovery Table, an ACPI table. 91 92 Note: QTG 0 is the only one currently supported in QEMU. 93 94 CXL Host Bridge (CXL HB) 95 ~~~~~~~~~~~~~~~~~~~~~~~~ 96 A CXL host bridge is similar to the PCIe equivalent, but with a 97 specification defined register interface called CXL Host Bridge 98 Component Registers (CHBCR). The location of this CHBCR MMIO 99 space is described to system software via a CXL Host Bridge 100 Structure (CHBS) in the CEDT ACPI table. The actual interfaces 101 are identical to those used for other parts of the CXL hierarchy 102 as CXL Component Registers in PCI BARs. 103 104 Interfaces provided include: 105 106 * Configuration of HDM Decoders to route CXL Memory accesses with 107 a particularly Host Physical Address range to the target port 108 below which the CXL device servicing that address lies. This 109 may be a mapping to a single Root Port (RP) or across a set of 110 target RPs. 111 112 CXL Root Ports (CXL RP) 113 ~~~~~~~~~~~~~~~~~~~~~~~ 114 A CXL Root Port serves the same purpose as a PCIe Root Port. 115 There are a number of CXL specific Designated Vendor Specific 116 Extended Capabilities (DVSEC) in PCIe Configuration Space 117 and associated component register access via PCI bars. 118 119 CXL Switch 120 ~~~~~~~~~~ 121 Here we consider a simple CXL switch with only a single 122 virtual hierarchy. Whilst more complex devices exist, their 123 visibility to a particular host is generally the same as for 124 a simple switch design. Hosts often have no awareness 125 of complex rerouting and device pooling, they simply see 126 devices being hot added or hot removed. 127 128 A CXL switch has a similar architecture to those in PCIe, 129 with a single upstream port, internal PCI bus and multiple 130 downstream ports. 131 132 Both the CXL upstream and downstream ports have CXL specific 133 DVSECs in configuration space, and component registers in PCI 134 BARs. The Upstream Port has the configuration interfaces for 135 the HDM decoders which route incoming memory accesses to the 136 appropriate downstream port. 137 138 A CXL switch is created in a similar fashion to PCI switches 139 by creating an upstream port (cxl-upstream) and a number of 140 downstream ports on the internal switch bus (cxl-downstream). 141 142 CXL Memory Devices - Type 3 143 ~~~~~~~~~~~~~~~~~~~~~~~~~~~ 144 CXL type 3 devices use a PCI class code and are intended to be supported 145 by a generic operating system driver. They have HDM decoders 146 though in these EP devices, the decoder is responsible not for 147 routing but for translation of the incoming host physical address (HPA) 148 into a Device Physical Address (DPA). 149 150 CXL Memory Interleave 151 --------------------- 152 To understand the interaction of different CXL hardware components which 153 are emulated in QEMU, let us consider a memory read in a fully configured 154 CXL topology. Note that system software is responsible for configuration 155 of all components with the exception of the CFMWs. System software is 156 responsible for allocating appropriate ranges from within the CFMWs 157 and exposing those via normal memory configurations as would be done 158 for system RAM. 159 160 Example system Topology. x marks the match in each decoder level:: 161 162 |<------------------SYSTEM PHYSICAL ADDRESS MAP (1)----------------->| 163 | __________ __________________________________ __________ | 164 | | | | | | | | 165 | | CFMW 0 | | CXL Fixed Memory Window 1 | | CFMW 1 | | 166 | | HB0 only | | Configured to interleave memory | | HB1 only | | 167 | | | | memory accesses across HB0/HB1 | | | | 168 | |__________| |_____x____________________________| |__________| | 169 | | | | 170 | | | | 171 | | | | 172 | Interleave Decoder | | 173 | Matches this HB | | 174 \_____________| |_____________/ 175 __________|__________ _____|_______________ 176 | | | | 177 (2) | CXL HB 0 | | CXL HB 1 | 178 | HB IntLv Decoders | | HB IntLv Decoders | 179 | PCI/CXL Root Bus 0c | | PCI/CXL Root Bus 0d | 180 | | | | 181 |___x_________________| |_____________________| 182 | | | | 183 | | | | 184 A HB 0 HDM Decoder | | | 185 matches this Port | | | 186 | | | | 187 ___________|___ __________|__ __|_________ ___|_________ 188 (3)| Root Port 0 | | Root Port 1 | | Root Port 2| | Root Port 3 | 189 | Appears in | | Appears in | | Appears in | | Appear in | 190 | PCI topology | | PCI Topology| | PCI Topo | | PCI Topo | 191 | As 0c:00.0 | | as 0c:01.0 | | as de:00.0 | | as de:01.0 | 192 |_______________| |_____________| |____________| |_____________| 193 | | | | 194 | | | | 195 _____|_________ ______|______ ______|_____ ______|_______ 196 (4)| x | | | | | | | 197 | CXL Type3 0 | | CXL Type3 1 | | CXL type3 2| | CLX Type 3 3 | 198 | | | | | | | | 199 | PMEM0(Vol LSA)| | PMEM1 (...) | | PMEM2 (...)| | PMEM3 (...) | 200 | Decoder to go | | | | | | | 201 | from host PA | | PCI 0e:00.0 | | PCI df:00.0| | PCI e0:00.0 | 202 | to device PA | | | | | | | 203 | PCI as 0d:00.0| | | | | | | 204 |_______________| |_____________| |____________| |______________| 205 206 Notes: 207 208 (1) **3 CXL Fixed Memory Windows (CFMW)** corresponding to different 209 ranges of the system physical address map. Each CFMW has 210 particular interleave setup across the CXL Host Bridges (HB) 211 CFMW0 provides uninterleaved access to HB0, CFW2 provides 212 uninterleaved access to HB1. CFW1 provides interleaved memory access 213 across HB0 and HB1. 214 215 (2) **Two CXL Host Bridges**. Each of these has 2 CXL Root Ports and 216 programmable HDM decoders to route memory accesses either to 217 a single port or interleave them across multiple ports. 218 A complex configuration here, might be to use the following HDM 219 decoders in HB0. HDM0 routes CFMW0 requests to RP0 and hence 220 part of CXL Type3 0. HDM1 routes CFMW0 requests from a 221 different region of the CFMW0 PA range to RP2 and hence part 222 of CXL Type 3 1. HDM2 routes yet another PA range from within 223 CFMW0 to be interleaved across RP0 and RP1, providing 2 way 224 interleave of part of the memory provided by CXL Type3 0 and 225 CXL Type 3 1. HDM3 routes those interleaved accesses from 226 CFMW1 that target HB0 to RP 0 and another part of the memory of 227 CXL Type 3 0 (as part of a 2 way interleave at the system level 228 across for example CXL Type3 0 and CXL Type3 2. 229 HDM4 is used to enable system wide 4 way interleave across all 230 the present CXL type3 devices, by interleaving those (interleaved) 231 requests that HB0 receives from from CFMW1 across RP 0 and 232 RP 1 and hence to yet more regions of the memory of the 233 attached Type3 devices. Note this is a representative subset 234 of the full range of possible HDM decoder configurations in this 235 topology. 236 237 (3) **Four CXL Root Ports.** In this case the CXL Type 3 devices are 238 directly attached to these ports. 239 240 (4) **Four CXL Type3 memory expansion devices.** These will each have 241 HDM decoders, but in this case rather than performing interleave 242 they will take the Host Physical Addresses of accesses and map 243 them to their own local Device Physical Address Space (DPA). 244 245 Example topology involving a switch:: 246 247 |<------------------SYSTEM PHYSICAL ADDRESS MAP (1)----------------->| 248 | __________ __________________________________ __________ | 249 | | | | | | | | 250 | | CFMW 0 | | CXL Fixed Memory Window 1 | | CFMW 1 | | 251 | | HB0 only | | Configured to interleave memory | | HB1 only | | 252 | | | | memory accesses across HB0/HB1 | | | | 253 | |____x_____| |__________________________________| |__________| | 254 | | | | 255 | | | | 256 | | | 257 Interleave Decoder | | | 258 Matches this HB | | | 259 \_____________| |_____________/ 260 __________|__________ _____|_______________ 261 | | | | 262 | CXL HB 0 | | CXL HB 1 | 263 | HB IntLv Decoders | | HB IntLv Decoders | 264 | PCI/CXL Root Bus 0c | | PCI/CXL Root Bus 0d | 265 | | | | 266 |___x_________________| |_____________________| 267 | | | | 268 | 269 A HB 0 HDM Decoder 270 matches this Port 271 ___________|___ 272 | Root Port 0 | 273 | Appears in | 274 | PCI topology | 275 | As 0c:00.0 | 276 |___________x___| 277 | 278 | 279 \_____________________ 280 | 281 | 282 --------------------------------------------------- 283 | Switch 0 USP as PCI 0d:00.0 | 284 | USP has HDM decoder which direct traffic to | 285 | appropriate downstream port | 286 | Switch BUS appears as 0e | 287 |x__________________________________________________| 288 | | | | 289 | | | | 290 _____|_________ ______|______ ______|_____ ______|_______ 291 (4)| x | | | | | | | 292 | CXL Type3 0 | | CXL Type3 1 | | CXL type3 2| | CLX Type 3 3 | 293 | | | | | | | | 294 | PMEM0(Vol LSA)| | PMEM1 (...) | | PMEM2 (...)| | PMEM3 (...) | 295 | Decoder to go | | | | | | | 296 | from host PA | | PCI 10:00.0 | | PCI 11:00.0| | PCI 12:00.0 | 297 | to device PA | | | | | | | 298 | PCI as 0f:00.0| | | | | | | 299 |_______________| |_____________| |____________| |______________| 300 301 Example command lines 302 --------------------- 303 A very simple setup with just one directly attached CXL Type 3 device:: 304 305 qemu-system-aarch64 -M virt,gic-version=3,cxl=on -m 4g,maxmem=8G,slots=8 -cpu max \ 306 ... 307 -object memory-backend-file,id=cxl-mem1,share=on,mem-path=/tmp/cxltest.raw,size=256M \ 308 -object memory-backend-file,id=cxl-lsa1,share=on,mem-path=/tmp/lsa.raw,size=256M \ 309 -device pxb-cxl,bus_nr=12,bus=pcie.0,id=cxl.1 \ 310 -device cxl-rp,port=0,bus=cxl.1,id=root_port13,chassis=0,slot=2 \ 311 -device cxl-type3,bus=root_port13,memdev=cxl-mem1,lsa=cxl-lsa1,id=cxl-pmem0 \ 312 -M cxl-fmw.0.targets.0=cxl.1,cxl-fmw.0.size=4G 313 314 A setup suitable for 4 way interleave. Only one fixed window provided, to enable 2 way 315 interleave across 2 CXL host bridges. Each host bridge has 2 CXL Root Ports, with 316 the CXL Type3 device directly attached (no switches).:: 317 318 qemu-system-aarch64 -M virt,gic-version=3,cxl=on -m 4g,maxmem=8G,slots=8 -cpu max \ 319 ... 320 -object memory-backend-file,id=cxl-mem1,share=on,mem-path=/tmp/cxltest.raw,size=256M \ 321 -object memory-backend-file,id=cxl-mem2,share=on,mem-path=/tmp/cxltest2.raw,size=256M \ 322 -object memory-backend-file,id=cxl-mem3,share=on,mem-path=/tmp/cxltest3.raw,size=256M \ 323 -object memory-backend-file,id=cxl-mem4,share=on,mem-path=/tmp/cxltest4.raw,size=256M \ 324 -object memory-backend-file,id=cxl-lsa1,share=on,mem-path=/tmp/lsa.raw,size=256M \ 325 -object memory-backend-file,id=cxl-lsa2,share=on,mem-path=/tmp/lsa2.raw,size=256M \ 326 -object memory-backend-file,id=cxl-lsa3,share=on,mem-path=/tmp/lsa3.raw,size=256M \ 327 -object memory-backend-file,id=cxl-lsa4,share=on,mem-path=/tmp/lsa4.raw,size=256M \ 328 -device pxb-cxl,bus_nr=12,bus=pcie.0,id=cxl.1 \ 329 -device pxb-cxl,bus_nr=222,bus=pcie.0,id=cxl.2 \ 330 -device cxl-rp,port=0,bus=cxl.1,id=root_port13,chassis=0,slot=2 \ 331 -device cxl-type3,bus=root_port13,memdev=cxl-mem1,lsa=cxl-lsa1,id=cxl-pmem0 \ 332 -device cxl-rp,port=1,bus=cxl.1,id=root_port14,chassis=0,slot=3 \ 333 -device cxl-type3,bus=root_port14,memdev=cxl-mem2,lsa=cxl-lsa2,id=cxl-pmem1 \ 334 -device cxl-rp,port=0,bus=cxl.2,id=root_port15,chassis=0,slot=5 \ 335 -device cxl-type3,bus=root_port15,memdev=cxl-mem3,lsa=cxl-lsa3,id=cxl-pmem2 \ 336 -device cxl-rp,port=1,bus=cxl.2,id=root_port16,chassis=0,slot=6 \ 337 -device cxl-type3,bus=root_port16,memdev=cxl-mem4,lsa=cxl-lsa4,id=cxl-pmem3 \ 338 -M cxl-fmw.0.targets.0=cxl.1,cxl-fmw.0.targets.1=cxl.2,cxl-fmw.0.size=4G,cxl-fmw.0.interleave-granularity=8k 339 340 An example of 4 devices below a switch suitable for 1, 2 or 4 way interleave:: 341 342 qemu-system-aarch64 -M virt,gic-version=3,cxl=on -m 4g,maxmem=8G,slots=8 -cpu max \ 343 ... 344 -object memory-backend-file,id=cxl-mem0,share=on,mem-path=/tmp/cxltest.raw,size=256M \ 345 -object memory-backend-file,id=cxl-mem1,share=on,mem-path=/tmp/cxltest1.raw,size=256M \ 346 -object memory-backend-file,id=cxl-mem2,share=on,mem-path=/tmp/cxltest2.raw,size=256M \ 347 -object memory-backend-file,id=cxl-mem3,share=on,mem-path=/tmp/cxltest3.raw,size=256M \ 348 -object memory-backend-file,id=cxl-lsa0,share=on,mem-path=/tmp/lsa0.raw,size=256M \ 349 -object memory-backend-file,id=cxl-lsa1,share=on,mem-path=/tmp/lsa1.raw,size=256M \ 350 -object memory-backend-file,id=cxl-lsa2,share=on,mem-path=/tmp/lsa2.raw,size=256M \ 351 -object memory-backend-file,id=cxl-lsa3,share=on,mem-path=/tmp/lsa3.raw,size=256M \ 352 -device pxb-cxl,bus_nr=12,bus=pcie.0,id=cxl.1 \ 353 -device cxl-rp,port=0,bus=cxl.1,id=root_port0,chassis=0,slot=0 \ 354 -device cxl-rp,port=1,bus=cxl.1,id=root_port1,chassis=0,slot=1 \ 355 -device cxl-upstream,bus=root_port0,id=us0 \ 356 -device cxl-downstream,port=0,bus=us0,id=swport0,chassis=0,slot=4 \ 357 -device cxl-type3,bus=swport0,memdev=cxl-mem0,lsa=cxl-lsa0,id=cxl-pmem0,size=256M \ 358 -device cxl-downstream,port=1,bus=us0,id=swport1,chassis=0,slot=5 \ 359 -device cxl-type3,bus=swport1,memdev=cxl-mem1,lsa=cxl-lsa1,id=cxl-pmem1,size=256M \ 360 -device cxl-downstream,port=2,bus=us0,id=swport2,chassis=0,slot=6 \ 361 -device cxl-type3,bus=swport2,memdev=cxl-mem2,lsa=cxl-lsa2,id=cxl-pmem2,size=256M \ 362 -device cxl-downstream,port=3,bus=us0,id=swport3,chassis=0,slot=7 \ 363 -device cxl-type3,bus=swport3,memdev=cxl-mem3,lsa=cxl-lsa3,id=cxl-pmem3,size=256M \ 364 -M cxl-fmw.0.targets.0=cxl.1,cxl-fmw.0.size=4G,cxl-fmw.0.interleave-granularity=4k 365 366 Kernel Configuration Options 367 ---------------------------- 368 369 In Linux 5.18 the following options are necessary to make use of 370 OS management of CXL memory devices as described here. 371 372 * CONFIG_CXL_BUS 373 * CONFIG_CXL_PCI 374 * CONFIG_CXL_ACPI 375 * CONFIG_CXL_PMEM 376 * CONFIG_CXL_MEM 377 * CONFIG_CXL_PORT 378 * CONFIG_CXL_REGION 379 380 References 381 ---------- 382 383 - Consortium website for specifications etc: 384 http://www.computeexpresslink.org 385 - Compute Express link Revision 2 specification, October 2020 386 - CEDT CFMWS & QTG _DSM ECN May 2021 387