| Message ID | b1b9e70db9d08364040436154cf0cf2f4600da49.1479788404.git.panand@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Pratyush, On 11/21/2016 08:32 PM, Pratyush Anand wrote: > When "enable-dcache" is passed to the kexec() command line, kexec-tools > passes this information to purgatory, which in turn enables cache during > sha-256 verification. What's the point of this enable-dcache option? Why not just always enable the cache if we can? -Geoff
Hi Geoff, On Wednesday 23 November 2016 12:27 AM, Geoff Levand wrote: > Hi Pratyush, > > On 11/21/2016 08:32 PM, Pratyush Anand wrote: >> When "enable-dcache" is passed to the kexec() command line, kexec-tools >> passes this information to purgatory, which in turn enables cache during >> sha-256 verification. > > What's the point of this enable-dcache option? Why not just > always enable the cache if we can? As I have written in changelog of patch 1/2 "We are supporting only 4K and 64K page sizes. This code will not work if a hardware is not supporting at least one of these page sizes. Therefore, D-cache is disabled by default and enabled only when "enable-dcache" is passed to the kexec()." Although this is very unlikely that a hardware will support only 16K page sizes, however it is possible. Therefore, its better to keep it disabled by default. ~Pratyush
On 11/23/16 at 07:16am, Pratyush Anand wrote: > Hi Geoff, > > On Wednesday 23 November 2016 12:27 AM, Geoff Levand wrote: > > Hi Pratyush, > > > > On 11/21/2016 08:32 PM, Pratyush Anand wrote: > > > When "enable-dcache" is passed to the kexec() command line, kexec-tools > > > passes this information to purgatory, which in turn enables cache during > > > sha-256 verification. > > > > What's the point of this enable-dcache option? Why not just > > always enable the cache if we can? > > As I have written in changelog of patch 1/2 > > "We are supporting only 4K and 64K page sizes. This code will not work if a > hardware is not supporting at least one of these page sizes. Therefore, > D-cache is disabled by default and enabled only when "enable-dcache" is > passed to the kexec()." > > > Although this is very unlikely that a hardware will support only 16K page > sizes, however it is possible. Therefore, its better to keep it disabled by > default. If it is *unlikely* it could be better to make it as default and add a --disable-dcache instead. > > ~Pratyush > > _______________________________________________ > kexec mailing list > kexec@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/kexec Thanks Dave
On Wednesday 23 November 2016 07:33 AM, Dave Young wrote: >> Although this is very unlikely that a hardware will support only 16K page >> > sizes, however it is possible. Therefore, its better to keep it disabled by >> > default. > If it is *unlikely* it could be better to make it as default and add a > --disable-dcache instead. > I think, I can do that. ~Pratyush
On Wed, Nov 23, 2016 at 07:41:52AM +0530, Pratyush Anand wrote: > > > On Wednesday 23 November 2016 07:33 AM, Dave Young wrote: > >>Although this is very unlikely that a hardware will support only 16K page > >>> sizes, however it is possible. Therefore, its better to keep it disabled by > >>> default. > >If it is *unlikely* it could be better to make it as default and add a > >--disable-dcache instead. > > > > I think, I can do that. Can this be detected at run-time? It sounds like it will be painful if on some setups the default doesn't work.
Hi Simon, Thanks for your review comment. On Wed, Nov 23, 2016 at 1:38 PM, Simon Horman <horms@verge.net.au> wrote: > On Wed, Nov 23, 2016 at 07:41:52AM +0530, Pratyush Anand wrote: >> >> >> On Wednesday 23 November 2016 07:33 AM, Dave Young wrote: >> >>Although this is very unlikely that a hardware will support only 16K page >> >>> sizes, however it is possible. Therefore, its better to keep it disabled by >> >>> default. >> >If it is *unlikely* it could be better to make it as default and add a >> >--disable-dcache instead. >> > >> >> I think, I can do that. > > Can this be detected at run-time? Thats doable. OK, so if everyone agrees then I can send a V2 where neither --enable-dcache nor --disable-dcache will be used. It will enable dcache if 4K/64K page is supported and will do nothing otherwise. ~Pratyush
diff --git a/kexec/arch/arm64/include/arch/options.h b/kexec/arch/arm64/include/arch/options.h index a17d933e396b..3e76ff04d6c1 100644 --- a/kexec/arch/arm64/include/arch/options.h +++ b/kexec/arch/arm64/include/arch/options.h @@ -5,13 +5,15 @@ #define OPT_DTB ((OPT_MAX)+1) #define OPT_INITRD ((OPT_MAX)+2) #define OPT_REUSE_CMDLINE ((OPT_MAX)+3) -#define OPT_ARCH_MAX ((OPT_MAX)+4) +#define OPT_ENABLE_DCACHE ((OPT_MAX)+4) +#define OPT_ARCH_MAX ((OPT_MAX)+5) #define KEXEC_ARCH_OPTIONS \ KEXEC_OPTIONS \ { "append", 1, NULL, OPT_APPEND }, \ { "command-line", 1, NULL, OPT_APPEND }, \ { "dtb", 1, NULL, OPT_DTB }, \ + { "enable-dcache", 0, NULL, OPT_ENABLE_DCACHE }, \ { "initrd", 1, NULL, OPT_INITRD }, \ { "ramdisk", 1, NULL, OPT_INITRD }, \ { "reuse-cmdline", 0, NULL, OPT_REUSE_CMDLINE }, \ @@ -24,6 +26,7 @@ static const char arm64_opts_usage[] __attribute__ ((unused)) = " --append=STRING Set the kernel command line to STRING.\n" " --command-line=STRING Set the kernel command line to STRING.\n" " --dtb=FILE Use FILE as the device tree blob.\n" +" --enable-dcache Enable D-Cache in Purgatory for faster SHA verification.\n" " --initrd=FILE Use FILE as the kernel initial ramdisk.\n" " --ramdisk=FILE Use FILE as the kernel initial ramdisk.\n" " --reuse-cmdline Use kernel command line from running system.\n"; @@ -32,6 +35,7 @@ struct arm64_opts { const char *command_line; const char *dtb; const char *initrd; + uint8_t enable_dcache; }; extern struct arm64_opts arm64_opts; diff --git a/kexec/arch/arm64/include/types.h b/kexec/arch/arm64/include/types.h new file mode 100644 index 000000000000..08f833a6d585 --- /dev/null +++ b/kexec/arch/arm64/include/types.h @@ -0,0 +1,16 @@ +#ifndef _TYPES_H_ +#define _TYPES_H_ + +#define min(x,y) ({ \ + typeof(x) _x = (x); \ + typeof(y) _y = (y); \ + (void) (&_x == &_y); \ + _x < _y ? _x : _y; }) + +#define max(x,y) ({ \ + typeof(x) _x = (x); \ + typeof(y) _y = (y); \ + (void) (&_x == &_y); \ + _x > _y ? _x : _y; }) + +#endif /* _TYPES_H_ */ diff --git a/kexec/arch/arm64/kexec-arm64.c b/kexec/arch/arm64/kexec-arm64.c index 288548f49304..b54d1b5304f6 100644 --- a/kexec/arch/arm64/kexec-arm64.c +++ b/kexec/arch/arm64/kexec-arm64.c @@ -23,6 +23,7 @@ #include "fs2dt.h" #include "kexec-syscall.h" #include "arch/options.h" +#include "types.h" /* Global varables the core kexec routines expect. */ @@ -130,6 +131,9 @@ int arch_process_options(int argc, char **argv) case OPT_PANIC: die("load-panic (-p) not supported"); break; + case OPT_ENABLE_DCACHE: + arm64_opts.enable_dcache = 1; + break; default: break; /* Ignore core and unknown options. */ } @@ -323,10 +327,13 @@ unsigned long arm64_locate_kernel_segment(struct kexec_info *info) int arm64_load_other_segments(struct kexec_info *info, unsigned long image_base) { - int result; + int result, i; unsigned long dtb_base; unsigned long hole_min; unsigned long hole_max; + unsigned long arm64_ram_start = -1; + unsigned long arm64_ram_end = 0; + uint8_t purgatory_enable_dcache; char *initrd_buf = NULL; struct dtb dtb; char command_line[COMMAND_LINE_SIZE] = ""; @@ -337,6 +344,8 @@ int arm64_load_other_segments(struct kexec_info *info, command_line[sizeof(command_line) - 1] = 0; } + purgatory_enable_dcache = arm64_opts.enable_dcache; + if (arm64_opts.dtb) { dtb.name = "dtb_user"; dtb.buf = slurp_file(arm64_opts.dtb, &dtb.size); @@ -419,8 +428,22 @@ int arm64_load_other_segments(struct kexec_info *info, elf_rel_set_symbol(&info->rhdr, "arm64_kernel_entry", &image_base, sizeof(image_base)); + elf_rel_set_symbol(&info->rhdr, "arm64_enable_dcache", + &purgatory_enable_dcache, sizeof(purgatory_enable_dcache)); + elf_rel_set_symbol(&info->rhdr, "arm64_dtb_addr", &dtb_base, sizeof(dtb_base)); + for (i = 0; i < info->nr_segments; i++) { + arm64_ram_start = min(arm64_ram_start, + (unsigned long)info->segment[i].mem); + arm64_ram_end = max(arm64_ram_end, + ((unsigned long)info->segment[i].mem + + info->segment[i].memsz)); + } + elf_rel_set_symbol(&info->rhdr, "arm64_ram_start", + &arm64_ram_start, sizeof(arm64_ram_start)); + elf_rel_set_symbol(&info->rhdr, "arm64_ram_end", + &arm64_ram_end, sizeof(arm64_ram_end)); return 0; } diff --git a/purgatory/arch/arm64/purgatory-arm64.c b/purgatory/arch/arm64/purgatory-arm64.c index fe50fcf8ebc3..6d61dcbce9ac 100644 --- a/purgatory/arch/arm64/purgatory-arm64.c +++ b/purgatory/arch/arm64/purgatory-arm64.c @@ -4,6 +4,13 @@ #include <stdint.h> #include <purgatory.h> +#include "cache.h" + +/* Symbols set by kexec. */ + +uint8_t arm64_enable_dcache __attribute__ ((section ("data"))); +uint64_t arm64_ram_start __attribute__ ((section ("data"))); +uint64_t arm64_ram_end __attribute__ ((section ("data"))); void putchar(int ch) { @@ -12,8 +19,12 @@ void putchar(int ch) void post_verification_setup_arch(void) { + if (arm64_enable_dcache) + disable_dcache(arm64_ram_start, arm64_ram_end); } void setup_arch(void) { + if (arm64_enable_dcache) + enable_dcache(arm64_ram_start, arm64_ram_end, 0); }
When "enable-dcache" is passed to the kexec() command line, kexec-tools passes this information to purgatory, which in turn enables cache during sha-256 verification. RAM boundary which includes all the sections is needed for creating identity page mapping and to enable d-cache for those areas. Therefore these informations are passed to purgatory as well. Signed-off-by: Pratyush Anand <panand@redhat.com> --- kexec/arch/arm64/include/arch/options.h | 6 +++++- kexec/arch/arm64/include/types.h | 16 ++++++++++++++++ kexec/arch/arm64/kexec-arm64.c | 25 ++++++++++++++++++++++++- purgatory/arch/arm64/purgatory-arm64.c | 11 +++++++++++ 4 files changed, 56 insertions(+), 2 deletions(-) create mode 100644 kexec/arch/arm64/include/types.h