[sw/silicon_creator] ROM digest and chip_info #14892

alphan · 2022-09-13T03:16:58Z

Noticed while working on #14884

The top eight words in ROM (by logical address) are interpreted as a 256-bit expected hash. Unlike the rest of ROM, their data is not stored scrambled, so the expected hash can be read directly.

Do we need to store chip_info in ROM?
If yes, we need to reserve 256 bits at the top of ROM for its digest. See test_rom.ld.

@moidx @cfrantz @tjaychen

The text was updated successfully, but these errors were encountered:

timothytrippel · 2022-09-13T05:07:27Z

Noticed this as well while working on #14890 :) ... was just about to file an issue thanks @alphan !

tjaychen · 2022-09-13T14:56:47Z

sorry i think i'm a little confused. So are we saying today during build, we insert chip_info into the top 128B, but during the rom_ctrl scramble we end up overriding it with the digest..? does that mean chip info doesn't print out correctly today at all?

tjaychen · 2022-09-13T14:57:19Z

ooo wait i see we subtract the digest size in test_rom..

tjaychen · 2022-09-13T15:06:07Z

yeah it feels like this would be good to have in the real rom.
Since someone could look at the printout and reference back to the repo for a particular githash.

alphan · 2022-09-13T16:15:56Z

From ROM sync: Remove chip_info, use 32-bit truncated commit hash, print during shutdown. Use VER: for the prefix.

andreaskurth · 2023-02-28T13:51:07Z

Assigning to M2.5 with Priority:P2 Priority: medium because ROM will be frozen after that. Please feel free to defer post M2.5 if we should not implement this for this release.

(I came across this while triaging rom_ctrl.)

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map Issue lowRISC#14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin For example, here's the end of the ROM from the previous command. Note that it contains 32 bits of a git commit hash starting at 0x7f80, followed by a human-readable timestamp. ... 00007f70: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 00007f80: e3fe 72cd 3230 3233 2d30 342d 3230 2031 ..r.2023-04-20 1 00007f90: 363a 3533 3a34 3800 6:53:48. Issue lowRISC#14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin For example, here's the end of the ROM from the previous command. Note that it contains 32 bits of a git commit hash starting at 0x7f80, followed by a human-readable timestamp. ... 00007f70: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 00007f80: 0000 0000 633e 05dd e3fe 72cd 3230 3233 ....c>....r.2023 00007f90: 2d30 342d 3231 2031 353a 3237 3a30 3100 -04-21 15:27:01. Issue lowRISC#14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin For example, here's the end of the ROM from the previous command. Note that it contains 32 bits of a git commit hash starting at 0x7f80, followed by a human-readable timestamp. 00007f80: 0000 0000 0000 0000 633e 05dd e3fe 72cd ........c>....r. 00007f90: 3230 3233 2d30 342d 3231 2031 353a 3237 2023-04-21 15:27 00007fa0: 3a30 3100 0000 0000 :01..... Issue lowRISC#14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin For example, here's the end of the ROM from the previous command. Note that it contains 32 bits of a git commit hash starting at 0x7f80, followed by a human-readable timestamp. 00007f70: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 00007f80: 633e 05dd e3fe 72cd 0000 0000 3230 3233 c>....r.....2023 00007f90: 2d30 342d 3235 2031 353a 3133 3a32 3900 -04-25 15:13:29. Issue lowRISC#14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin For example, here's the end of the ROM from the previous command. Note that it contains 32 bits of a git commit hash starting at 0x7f80, followed by a human-readable timestamp. 00007f70: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 00007f80: 9461 ce0e c828 0d91 0000 0000 0000 0000 .a...(.......... Issue lowRISC#14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin For example, here's the end of the ROM from the previous command. At 0x7f80, there are 64 bits of a Git commit hash (e20871e1cf102bf6). 00007f70: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 00007f80: f62b 10cf e171 08e2 .+...q.. Issue lowRISC#14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

This commit adds a new `chip_info_t` struct that contains info about the ROM's provenance. The ROM now prints the Git commit hash on shutdown after the "VER:" prefix. The ROM linker script was already configured to place the .chip_info section at the top of ROM, but I'm not confident that it worked as expected. The autogenerated header contained a static constant, so I think a copy would be inlined at some arbitrary location into any compilation unit that used it, rather than being placed by the linker script. To verify that the ROM prints the Git commit hash on shutdown: ./bazelisk.sh test --test_output=streamed \ //sw/device/silicon_creator/rom/e2e:shutdown_output_dev_fpga_cw310_rom To verify that the .chip_info section contains data from chip_info.o: ./bazelisk.sh build-then 'less %s' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_map To see the value of `kChipInfo` at the end of ROM: ./bazelisk.sh build-then 'xxd %s | less' --config riscv32 \ //sw/device/silicon_creator/rom:rom_with_fake_keys_fpga_cw310_bin For example, here's the end of the ROM from the previous command. At 0x7f80, there are 64 bits of a Git commit hash (e20871e1cf102bf6). 00007f70: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 00007f80: f62b 10cf e171 08e2 .+...q.. Issue #14892 Signed-off-by: Dan McArdle <dmcardle@opentitan.org>

alphan added Component:Software Issue related to Software Type:Task Tasks, to-do list. SW:ROM ROM related issues labels Sep 13, 2022

alphan self-assigned this Sep 13, 2022

msfschaffner added this to the Project: M3 milestone Sep 30, 2022

andreaskurth modified the milestones: Discrete: M3, Discrete: M2.5 Feb 28, 2023

andreaskurth added the Priority:P2 Priority: medium label Feb 28, 2023

arunthomas assigned dmcardle and unassigned alphan Apr 17, 2023

dmcardle mentioned this issue Apr 20, 2023

[sw/silicon_creator] Add structured kChipInfo to ROM #18100

Merged

dmcardle mentioned this issue Apr 26, 2023

[sw/device] Discuss design of chip_info and lc_ctrl.HW_REV #18198

Closed

dmcardle closed this as completed in #18100 May 1, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[sw/silicon_creator] ROM digest and chip_info #14892

[sw/silicon_creator] ROM digest and chip_info #14892

alphan commented Sep 13, 2022

timothytrippel commented Sep 13, 2022

tjaychen commented Sep 13, 2022

tjaychen commented Sep 13, 2022

tjaychen commented Sep 13, 2022

alphan commented Sep 13, 2022 •

edited

andreaskurth commented Feb 28, 2023

[sw/silicon_creator] ROM digest and chip_info #14892

[sw/silicon_creator] ROM digest and chip_info #14892

Comments

alphan commented Sep 13, 2022

timothytrippel commented Sep 13, 2022

tjaychen commented Sep 13, 2022

tjaychen commented Sep 13, 2022

tjaychen commented Sep 13, 2022

alphan commented Sep 13, 2022 • edited

andreaskurth commented Feb 28, 2023

alphan commented Sep 13, 2022 •

edited