UEFI 2.8 defines an EFI_MEMORY_SP attribute bit to augment the interpretation of the EFI Memory Types as "reserved for a specific purpose". The proposed Linux behavior for specific purpose memory is that it is reserved for direct-access (device-dax) by default and not available for any kernel usage, not even as an OOM fallback. Later, through udev scripts or another init mechanism, these device-dax claimed ranges can be reconfigured and hot-added to the available System-RAM with a unique node identifier. This device-dax management scheme implements "soft" in the "soft reserved" designation by allowing some or all of the reservation to be recovered as typical memory. This policy can be disabled at compile-time with CONFIG_EFI_SOFT_RESERVE=n, or runtime with efi=nosoftreserve. As for this patch, define the common helpers to determine if the EFI_MEMORY_SP attribute should be honored. The determination needs to be made early to prevent the kernel from being loaded into soft-reserved memory, or otherwise allowing early allocations to land there. Follow-on changes are needed per architecture to leverage these helpers in their respective mem-init paths.
Unified · Split
| @@ -1168,7 +1168,8 @@ · 1167 unmodified lines | |||
| 1168 | Format: {"off" | "on" | "skip[mbr]"} | 1168 | Format: {"off" | "on" | "skip[mbr]"} |
| 1169 | 1169 | ||
| 1170 | efi= [EFI] | 1170 | efi= [EFI] |
| 1171 | Format: { "old_map", "nochunk", "noruntime", "debug" } | 1171 | Format: { "old_map", "nochunk", "noruntime", "debug", |
| 1172 | "nosoftreserve" } | ||
| 1172 | old_map [X86-64]: switch to the old ioremap-based EFI | 1173 | old_map [X86-64]: switch to the old ioremap-based EFI |
| 1173 | runtime services mapping. 32-bit still uses this one by | 1174 | runtime services mapping. 32-bit still uses this one by |
| 1174 | default. | 1175 | default. |
| @@ -1177,6 +1178,12 @@ · 2 unmodified lines | |||
| 1177 | firmware implementations. | 1178 | firmware implementations. |
| 1178 | noruntime : disable EFI runtime services support | 1179 | noruntime : disable EFI runtime services support |
| 1179 | debug: enable misc debug output | 1180 | debug: enable misc debug output |
| 1181 | nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) | ||
| 1182 | attribute may cause the kernel to reserve the | ||
| 1183 | memory range for a memory mapping driver to | ||
| 1184 | claim. Specify efi=nosoftreserve to disable this | ||
| 1185 | reservation and treat the memory by its base type | ||
| 1186 | (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). | ||
| 1180 | 1187 | ||
| 1181 | efi_no_storage_paranoia [EFI; X86] | 1188 | efi_no_storage_paranoia [EFI; X86] |
| 1182 | Using this parameter you can use more than 50% of | 1189 | Using this parameter you can use more than 50% of |
| @@ -75,6 +75,27 @@ config EFI_MAX_FAKE_MEM · 74 unmodified lines | |||
| 75 | Ranges can be set up to this value using comma-separated list. | 75 | Ranges can be set up to this value using comma-separated list. |
| 76 | The default value is 8. | 76 | The default value is 8. |
| 77 | 77 | ||
| 78 | config EFI_SOFT_RESERVE | ||
| 79 | bool "Reserve EFI Specific Purpose Memory" | ||
| 80 | depends on EFI && EFI_STUB && ACPI_HMAT | ||
| 81 | default ACPI_HMAT | ||
| 82 | help | ||
| 83 | On systems that have mixed performance classes of memory EFI | ||
| 84 | may indicate specific purpose memory with an attribute (See | ||
| 85 | EFI_MEMORY_SP in UEFI 2.8). A memory range tagged with this | ||
| 86 | attribute may have unique performance characteristics compared | ||
| 87 | to the system's general purpose "System RAM" pool. On the | ||
| 88 | expectation that such memory has application specific usage, | ||
| 89 | and its base EFI memory type is "conventional" answer Y to | ||
| 90 | arrange for the kernel to reserve it as a "Soft Reserved" | ||
| 91 | resource, and set aside for direct-access (device-dax) by | ||
| 92 | default. The memory range can later be optionally assigned to | ||
| 93 | the page allocator by system administrator policy via the | ||
| 94 | device-dax kmem facility. Say N to have the kernel treat this | ||
| 95 | memory as "System RAM" by default. | ||
| 96 | |||
| 97 | If unsure, say Y. | ||
| 98 | |||
| 78 | config EFI_PARAMS_FROM_FDT | 99 | config EFI_PARAMS_FROM_FDT |
| 79 | bool | 100 | bool |
| 80 | help | 101 | help |
| @@ -81,6 +81,11 @@ bool efi_runtime_disabled(void) · 80 unmodified lines | |||
| 81 | return disable_runtime; | 81 | return disable_runtime; |
| 82 | } | 82 | } |
| 83 | 83 | ||
| 84 | bool __pure __efi_soft_reserve_enabled(void) | ||
| 85 | { | ||
| 86 | return !efi_enabled(EFI_MEM_NO_SOFT_RESERVE); | ||
| 87 | } | ||
| 88 | |||
| 84 | static int __init parse_efi_cmdline(char *str) | 89 | static int __init parse_efi_cmdline(char *str) |
| 85 | { | 90 | { |
| 86 | if (!str) { | 91 | if (!str) { |
| @@ -94,6 +99,9 @@ static int __init parse_efi_cmdline(char *str) · 7 unmodified lines | |||
| 94 | if (parse_option_str(str, "noruntime")) | 99 | if (parse_option_str(str, "noruntime")) |
| 95 | disable_runtime = true; | 100 | disable_runtime = true; |
| 96 | 101 | ||
| 102 | if (parse_option_str(str, "nosoftreserve")) | ||
| 103 | set_bit(EFI_MEM_NO_SOFT_RESERVE, &efi.flags); | ||
| 104 | |||
| 97 | return 0; | 105 | return 0; |
| 98 | } | 106 | } |
| 99 | early_param("efi", parse_efi_cmdline); | 107 | early_param("efi", parse_efi_cmdline); |
| @@ -32,6 +32,7 @@ static unsigned long __chunk_size = EFI_READ_CHUNK_SIZE; · 31 unmodified lines | |||
| 32 | static int __section(.data) __nokaslr; | 32 | static int __section(.data) __nokaslr; |
| 33 | static int __section(.data) __quiet; | 33 | static int __section(.data) __quiet; |
| 34 | static int __section(.data) __novamap; | 34 | static int __section(.data) __novamap; |
| 35 | static bool __section(.data) efi_nosoftreserve; | ||
| 35 | 36 | ||
| 36 | int __pure nokaslr(void) | 37 | int __pure nokaslr(void) |
| 37 | { | 38 | { |
| @@ -45,6 +46,10 @@ int __pure novamap(void) · 7 unmodified lines | |||
| 45 | { | 46 | { |
| 46 | return __novamap; | 47 | return __novamap; |
| 47 | } | 48 | } |
| 49 | bool __pure __efi_soft_reserve_enabled(void) | ||
| 50 | { | ||
| 51 | return !efi_nosoftreserve; | ||
| 52 | } | ||
| 48 | 53 | ||
| 49 | #define EFI_MMAP_NR_SLACK_SLOTS 8 | 54 | #define EFI_MMAP_NR_SLACK_SLOTS 8 |
| 50 | 55 | ||
| @@ -211,6 +216,10 @@ again: · 160 unmodified lines | |||
| 211 | if (desc->type != EFI_CONVENTIONAL_MEMORY) | 216 | if (desc->type != EFI_CONVENTIONAL_MEMORY) |
| 212 | continue; | 217 | continue; |
| 213 | 218 | ||
| 219 | if (efi_soft_reserve_enabled() && | ||
| 220 | (desc->attribute & EFI_MEMORY_SP)) | ||
| 221 | continue; | ||
| 222 | |||
| 214 | if (desc->num_pages < nr_pages) | 223 | if (desc->num_pages < nr_pages) |
| 215 | continue; | 224 | continue; |
| 216 | 225 | ||
| @@ -305,6 +314,10 @@ efi_status_t efi_low_alloc_above(efi_system_table_t *sys_table_arg, · 88 unmodified lines | |||
| 305 | if (desc->type != EFI_CONVENTIONAL_MEMORY) | 314 | if (desc->type != EFI_CONVENTIONAL_MEMORY) |
| 306 | continue; | 315 | continue; |
| 307 | 316 | ||
| 317 | if (efi_soft_reserve_enabled() && | ||
| 318 | (desc->attribute & EFI_MEMORY_SP)) | ||
| 319 | continue; | ||
| 320 | |||
| 308 | if (desc->num_pages < nr_pages) | 321 | if (desc->num_pages < nr_pages) |
| 309 | continue; | 322 | continue; |
| 310 | 323 | ||
| @@ -484,6 +497,12 @@ efi_status_t efi_parse_options(char const *cmdline) · 173 unmodified lines | |||
| 484 | __novamap = 1; | 497 | __novamap = 1; |
| 485 | } | 498 | } |
| 486 | 499 | ||
| 500 | if (IS_ENABLED(CONFIG_EFI_SOFT_RESERVE) && | ||
| 501 | !strncmp(str, "nosoftreserve", 7)) { | ||
| 502 | str += strlen("nosoftreserve"); | ||
| 503 | efi_nosoftreserve = 1; | ||
| 504 | } | ||
| 505 | |||
| 487 | /* Group words together, delimited by "," */ | 506 | /* Group words together, delimited by "," */ |
| 488 | while (*str && *str != ' ' && *str != ',') | 507 | while (*str && *str != ' ' && *str != ',') |
| 489 | str++; | 508 | str++; |
| @@ -1202,6 +1202,7 @@ extern int __init efi_setup_pcdp_console(char *); · 1201 unmodified lines | |||
| 1202 | #define EFI_DBG 8 /* Print additional debug info at runtime */ | 1202 | #define EFI_DBG 8 /* Print additional debug info at runtime */ |
| 1203 | #define EFI_NX_PE_DATA 9 /* Can runtime data regions be mapped non-executable? */ | 1203 | #define EFI_NX_PE_DATA 9 /* Can runtime data regions be mapped non-executable? */ |
| 1204 | #define EFI_MEM_ATTR 10 /* Did firmware publish an EFI_MEMORY_ATTRIBUTES table? */ | 1204 | #define EFI_MEM_ATTR 10 /* Did firmware publish an EFI_MEMORY_ATTRIBUTES table? */ |
| 1205 | #define EFI_MEM_NO_SOFT_RESERVE 11 /* Is the kernel configured to ignore soft reservations? */ | ||
| 1205 | 1206 | ||
| 1206 | #ifdef CONFIG_EFI | 1207 | #ifdef CONFIG_EFI |
| 1207 | /* | 1208 | /* |
| @@ -1212,6 +1213,14 @@ static inline bool efi_enabled(int feature) · 4 unmodified lines | |||
| 1212 | return test_bit(feature, &efi.flags) != 0; | 1213 | return test_bit(feature, &efi.flags) != 0; |
| 1213 | } | 1214 | } |
| 1214 | extern void efi_reboot(enum reboot_mode reboot_mode, const char *__unused); | 1215 | extern void efi_reboot(enum reboot_mode reboot_mode, const char *__unused); |
| 1216 | |||
| 1217 | bool __pure __efi_soft_reserve_enabled(void); | ||
| 1218 | |||
| 1219 | static inline bool __pure efi_soft_reserve_enabled(void) | ||
| 1220 | { | ||
| 1221 | return IS_ENABLED(CONFIG_EFI_SOFT_RESERVE) | ||
| 1222 | && __efi_soft_reserve_enabled(); | ||
| 1223 | } | ||
| 1215 | #else | 1224 | #else |
| 1216 | static inline bool efi_enabled(int feature) | 1225 | static inline bool efi_enabled(int feature) |
| 1217 | { | 1226 | { |
| @@ -1225,6 +1234,11 @@ efi_capsule_pending(int *reset_type) · 7 unmodified lines | |||
| 1225 | { | 1234 | { |
| 1226 | return false; | 1235 | return false; |
| 1227 | } | 1236 | } |
| 1237 | |||
| 1238 | static inline bool efi_soft_reserve_enabled(void) | ||
| 1239 | { | ||
| 1240 | return false; | ||
| 1241 | } | ||
| 1228 | #endif | 1242 | #endif |
| 1229 | 1243 | ||
| 1230 | extern int efi_status_to_err(efi_status_t status); | 1244 | extern int efi_status_to_err(efi_status_t status); |