[kmac/dv] initial version of KMAC smoke test #4674

udinator · 2020-12-23T22:23:24Z

This PR adds in the initial draft of the KMAC smoke test, with some additional enhancement commits needed to support it.

The first commit adds a function to the str_utils_pkg to convert a bytestring into a string representation.

The second commit updates some wording in the testplan on the condition under which we manually squeeze more output data.

The third commit updates the kmac DPI package to add function name and customization string inputs to the CSHAKE and KMAC functions.

Finally, the last commit adds the actual testbench code to implement the smoke test.

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

weicaiyang · 2020-12-24T00:37:05Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+    csr_wr(.csr(ral.prefix_7), .value(prefix_arr[7]));
+    csr_wr(.csr(ral.prefix_8), .value(prefix_arr[8]));
+    csr_wr(.csr(ral.prefix_9), .value(prefix_arr[9]));
+    csr_wr(.csr(ral.prefix_10), .value(prefix_arr[10]));


nit: consider to use loop

you're right, I'll update it to use a similar loop to what you linked to below

weicaiyang · 2020-12-24T00:39:47Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+    csr_wr(.csr(ral.key_share0_0), .value(key_share0[0]));
+    csr_wr(.csr(ral.key_share0_1), .value(key_share0[1]));
+    csr_wr(.csr(ral.key_share0_2), .value(key_share0[2]));
+    csr_wr(.csr(ral.key_share0_3), .value(key_share0[3]));
+    csr_wr(.csr(ral.key_share0_4), .value(key_share0[4]));
+    csr_wr(.csr(ral.key_share0_5), .value(key_share0[5]));
+    csr_wr(.csr(ral.key_share0_6), .value(key_share0[6]));
+    csr_wr(.csr(ral.key_share0_7), .value(key_share0[7]));
+    csr_wr(.csr(ral.key_share0_8), .value(key_share0[8]));
+    csr_wr(.csr(ral.key_share0_9), .value(key_share0[9]));
+    csr_wr(.csr(ral.key_share0_10), .value(key_share0[10]));
+    csr_wr(.csr(ral.key_share0_11), .value(key_share0[11]));
+    csr_wr(.csr(ral.key_share0_12), .value(key_share0[12]));
+    csr_wr(.csr(ral.key_share0_13), .value(key_share0[13]));
+    csr_wr(.csr(ral.key_share0_14), .value(key_share0[14]));
+    csr_wr(.csr(ral.key_share0_15), .value(key_share0[15]));
+    // share 0
+    csr_wr(.csr(ral.key_share1_0), .value(key_share1[0]));
+    csr_wr(.csr(ral.key_share1_1), .value(key_share1[1]));
+    csr_wr(.csr(ral.key_share1_2), .value(key_share1[2]));
+    csr_wr(.csr(ral.key_share1_3), .value(key_share1[3]));
+    csr_wr(.csr(ral.key_share1_4), .value(key_share1[4]));
+    csr_wr(.csr(ral.key_share1_5), .value(key_share1[5]));
+    csr_wr(.csr(ral.key_share1_6), .value(key_share1[6]));
+    csr_wr(.csr(ral.key_share1_7), .value(key_share1[7]));
+    csr_wr(.csr(ral.key_share1_8), .value(key_share1[8]));
+    csr_wr(.csr(ral.key_share1_9), .value(key_share1[9]));
+    csr_wr(.csr(ral.key_share1_10), .value(key_share1[10]));
+    csr_wr(.csr(ral.key_share1_11), .value(key_share1[11]));
+    csr_wr(.csr(ral.key_share1_12), .value(key_share1[12]));
+    csr_wr(.csr(ral.key_share1_13), .value(key_share1[13]));
+    csr_wr(.csr(ral.key_share1_14), .value(key_share1[14]));
+    csr_wr(.csr(ral.key_share1_15), .value(key_share1[15]));


use loop like this

opentitan/hw/ip/keymgr/dv/env/seq_lib/keymgr_base_vseq.sv

Lines 170 to 176 in c00e37f

foreach (sw_share_output[i, j]) begin

string csr_name = $sformatf("sw_share%0d_output_%0d", i, j);

uvm_reg csr = ral.get_reg_by_name(csr_name);

csr_rd(.ptr(csr), .value(sw_share_output[i][j]));

`uvm_info(`gfn, $sformatf("%0s: 0x%0h", csr_name, sw_share_output[i][j]), UVM_HIGH)

end

ohh that's a very good point, I was trying to come up with some macro-based approach for this, but looping like what you do is definitely the best approach here.
Thanks for the suggestion!

weicaiyang · 2020-12-24T00:42:59Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+      // check that there is actually room in the fifo before we start writing anything
+      wait_fifo_has_capacity();
+      // can't use `DV_CHECK_RANDOMIZE_FATAL, so need a workaround
+      `DV_CHECK_FATAL(randomize(fifo_addr))


seems like we should use DV_CHECK_MEMBER_RANDOMIZE_FATAL

good point, will update this accordingly.

cindychip · 2021-01-05T18:44:38Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+  // Set `left_right` to 1'b1 for left encoding, and 1'b0 for right encoding
+  //
+  // Input `q` is assumed to be empty
+  function void encode(bit [2039:0] x, bit left_right, ref bit [7:0] q[$]);


Just a suggestion, maybe make 2040 a parameter?

sure, I can do that (would probably be clearer that way) ! thanks for the suggestion

cindychip · 2021-01-05T18:47:48Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+    // wait for all msgfifo accesses to complete
+    //
+    // TODO: is there a better way of doing this?
+    wait_no_outstanding_access();


Just a comment - I think the tl_access has a non_blocking option, https://github.com/lowRISC/opentitan/blob/master/hw/dv/sv/cip_lib/cip_base_vseq.sv#L102
Right now it defaults to blocking, so i believe there is no need to wait no outstanding.
But would be necessary if you set the non-blocking to 1.

yup you're absolutely right!
I intend to randomize blocking/nonblocking accesses in a future PR, so the call to wait_no_outstanding_access would become "relevant" then.
Let me comment out this function call for now with a TODO.

weicaiyang

LGTM

udinator · 2021-01-06T00:22:57Z

latest force push fixes failures in csr_hw_reset test.

hw/dv/sv/str_utils/str_utils_pkg.sv

sriyerg · 2021-01-11T21:35:31Z

hw/ip/kmac/dv/env/kmac_env_pkg.sv

+  // number of supported interrupts
+  parameter int KMAC_NUM_INTR = 3;


nit: An alternate approach is to add KmacNumIntrs as the last literal of kmac_intr_e (see UART for example), which has a couple of advantages.

what are some benefits of doing this? i'm not sure I fully understand how this would work

Enum literal is an immutable const, hence logically equivalent to a localparam

Looping through an enum is easy (use this literal as an exit condition). Without it, you will need to add a loop counter

No need to adjust the value of the parameter when interrupts are added / removed (forgetting to do that adds a small debug overhead)

Adds the ability to set an 'illegal' value to an enum variable

make sense!
I updated the enum to add this value at the very end.

sriyerg · 2021-01-11T21:38:45Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+    `uvm_info(`gfn, $sformatf("kmac_en: %0b", kmac_en), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("xof_en: %0b", xof_en), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("strength: %0s", strength.name()), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("keccak_block_size: %0d", keccak_block_size), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("hash_mode: %0s", hash_mode.name()), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("msg_endianness: %0b", msg_endian), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("state_endianness: %0b", state_endian), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("en_sideload: %0b", en_sideload), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("key_len: %0s", key_len.name()), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("output_len: %0d", output_len), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("fname_arr: %0p", fname_arr), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("fname: %0s", str_utils_pkg::bytes_to_str(fname_arr)), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("custom_str_arr: %0p", custom_str_arr), UVM_HIGH)
+    `uvm_info(`gfn, $sformatf("custom_str: %0s", str_utils_pkg::bytes_to_str(custom_str_arr)), UVM_HIGH)


Recommend adding a convert2string function and call it here instead.

good idea, thanks!

sriyerg · 2021-01-11T21:42:04Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+    custom_str_arr.size() == custom_str_len;
+  }
+
+  // constrains N and S to only be valid ASCII characters [A-Z,a-z]


Suggested change

// constrains N and S to only be valid ASCII characters [A-Z,a-z]

// constrains N and S to only be valid alphabet letters & space [A-Za-z ]

sriyerg · 2021-01-11T21:47:42Z

hw/ip/kmac/dv/env/seq_lib/kmac_base_vseq.sv

+  // general function to swap endianness of a byte array (such as msg/digest).
+  //
+  // This function preserves word ordering, and only endian-swaps the bytes in each word.
+  //
+  // e.g. if `arr[] = '{'h00, 'h01, 'h02, 'h03, 'h10, 'h11, 'h12, 'h13}`, the function will produce:
+  //      `'{'h03, 'h02, 'h01, 'h00, 'h13, 'h12, 'h11, 'h10}`
+  function void endian_swap(ref bit [7:0] arr[]);
+    arr = {<< byte {arr}};
+    arr = {<< 32 {arr}};
+  endfunction


Feel like we could move this to dv_utils_pkg.

sure, will do so. it seems like the dv_utils_pkg already has some endian functionality contained, so would be good to aggregate them all.

udinator · 2021-01-13T23:08:19Z

private CI failing is due to CSR test failures, #4515 resolves these issues so will need to wait until that gets merged.

This commit adds a `bytes_to_str` function in the `str_utils_pkg` that allows us to convert a bytestring back into string representation. Signed-off-by: Udi Jonnalagadda <udij@google.com>

This commit updates the KMAC testplan wording. Previously, it stated that we only manually squeeze more output data if the desired output length is greater than 1600 bits. This has been fixed to reflect that we only manually squeeze output data if the desired output length is greater than the keccak block size (168 or 136 bytes). Signed-off-by: Udi Jonnalagadda <udij@google.com>

This commit updates the KMAC DPI package to add function name and customization string inputs to the CSHAKE and KMAC functions. Signed-off-by: Udi Jonnalagadda <udij@google.com>

This commit adds the initial version of KMAC smoke test. There are some TODO items remaining that will be handled in the future, such as adding random sideload key injections, and moving most of the checking logic into the scoreboard once it is fleshed out. Signed-off-by: Udi Jonnalagadda <udij@google.com>

sriyerg

We can revisit the str bytes conversion later. The rest LGTM.

udinator · 2021-01-20T20:33:55Z

We can revisit the str bytes conversion later. The rest LGTM.

sgtm, thanks for the review!

udinator requested review from eunchan, sriyerg and weicaiyang as code owners December 23, 2020 22:23

udinator requested a review from cindychip December 23, 2020 22:24

udinator force-pushed the kmac_smoke branch from 535334d to cfb2805 Compare December 23, 2020 23:36

weicaiyang reviewed Dec 24, 2020

View reviewed changes

eunchan removed their request for review December 28, 2020 19:07

udinator force-pushed the kmac_smoke branch from cfb2805 to e1a11cb Compare January 5, 2021 18:20

cindychip reviewed Jan 5, 2021

View reviewed changes

udinator force-pushed the kmac_smoke branch from e1a11cb to 368ca4e Compare January 5, 2021 22:21

udinator requested a review from weicaiyang January 5, 2021 22:52

weicaiyang approved these changes Jan 5, 2021

View reviewed changes

udinator force-pushed the kmac_smoke branch from 368ca4e to 198ccc8 Compare January 6, 2021 00:22

sriyerg reviewed Jan 11, 2021

View reviewed changes

hw/dv/sv/str_utils/str_utils_pkg.sv Show resolved Hide resolved

sriyerg reviewed Jan 11, 2021

View reviewed changes

sriyerg reviewed Jan 12, 2021

View reviewed changes

udinator force-pushed the kmac_smoke branch from 198ccc8 to 1eaf0a5 Compare January 13, 2021 21:37

Udi Jonnalagadda added 3 commits January 14, 2021 10:48

[dv/str_utils_pkg] add byte_to_str function

9d7b546

This commit adds a `bytes_to_str` function in the `str_utils_pkg` that allows us to convert a bytestring back into string representation. Signed-off-by: Udi Jonnalagadda <udij@google.com>

[kmac/dv] update kmac DPI package

33d9a8a

This commit updates the KMAC DPI package to add function name and customization string inputs to the CSHAKE and KMAC functions. Signed-off-by: Udi Jonnalagadda <udij@google.com>

udinator force-pushed the kmac_smoke branch from 1eaf0a5 to 8cce2e3 Compare January 14, 2021 18:54

udinator force-pushed the kmac_smoke branch from 8cce2e3 to c9347e6 Compare January 15, 2021 00:15

sriyerg approved these changes Jan 20, 2021

View reviewed changes

udinator merged commit 7488dff into lowRISC:master Jan 20, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[kmac/dv] initial version of KMAC smoke test #4674

[kmac/dv] initial version of KMAC smoke test #4674

udinator commented Dec 23, 2020 •

edited

Loading

weicaiyang Dec 24, 2020

udinator Dec 24, 2020

weicaiyang Dec 24, 2020

udinator Dec 24, 2020

weicaiyang Dec 24, 2020

udinator Dec 24, 2020

cindychip Jan 5, 2021

udinator Jan 5, 2021

cindychip Jan 5, 2021

udinator Jan 5, 2021

weicaiyang left a comment

udinator commented Jan 6, 2021

sriyerg Jan 11, 2021

udinator Jan 12, 2021

sriyerg Jan 12, 2021

udinator Jan 12, 2021

sriyerg Jan 11, 2021

udinator Jan 12, 2021

sriyerg Jan 11, 2021

udinator Jan 12, 2021

sriyerg Jan 11, 2021

udinator Jan 12, 2021

udinator commented Jan 13, 2021

sriyerg left a comment

udinator commented Jan 20, 2021

	foreach (sw_share_output[i, j]) begin
	string csr_name = $sformatf("sw_share%0d_output_%0d", i, j);
	uvm_reg csr = ral.get_reg_by_name(csr_name);

	csr_rd(.ptr(csr), .value(sw_share_output[i][j]));
	`uvm_info(`gfn, $sformatf("%0s: 0x%0h", csr_name, sw_share_output[i][j]), UVM_HIGH)
	end

		// number of supported interrupts
		parameter int KMAC_NUM_INTR = 3;

	// constrains N and S to only be valid ASCII characters [A-Z,a-z]
	// constrains N and S to only be valid alphabet letters & space [A-Za-z ]

[kmac/dv] initial version of KMAC smoke test #4674

[kmac/dv] initial version of KMAC smoke test #4674

Conversation

udinator commented Dec 23, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

weicaiyang left a comment

Choose a reason for hiding this comment

udinator commented Jan 6, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

udinator commented Jan 13, 2021

sriyerg left a comment

Choose a reason for hiding this comment

udinator commented Jan 20, 2021

udinator commented Dec 23, 2020 •

edited

Loading