k3-binman ergonomics
cover k3-binman ergonomics
When trying to hook up use of a customer's own signing infrastructure for some am62x devices, I found it was more difficult than it should be to point at the right key(s). One has to patch in the keyfile= property in way too many places, and if you miss just one, the build succeeds using the generic custMpk.pem for the one you missed, but of course the resulting artifacts are unbootable.
Use the template mechanism that binman already has to allow one to only set the keyfile= in a single place. I cannot really imagine any case where one would not want to use the same key for signing all artifacts across the r5/a53 cores, but that is of course still possible, just as any downstreams that have explicit
&{/binman/tiboot3-am62x-hs-evm.bin/ti-secure-rom} {
keyfile = "foobar.pem";
};&{/binman/tiboot3-am62x-hs-fs-evm.bin/ti-secure-rom} {
keyfile = "foobar.pem";
};will continue to work.
Rasmus Villemoes (6):
binman: openssl: refactor creation of Distinguished Name section from
dict
binman: x509_cert: allow and parse distinguished-name subnode
binman: x509: fix CN emitted for basic x509 certificates k3-binman.dtsi: add keyfile_template node
k3-binman.dtsi: add empty distinguished_name_template k3-binman.dtsi: insert keyfile and distinguished name templates in all
ti-secure(-rom) nodes
arch/arm/dts/k3-binman.dtsi | 34 +++++++++++----- tools/binman/btool/openssl.py | 51 ++++++++---------------- tools/binman/etype/ti_secure.py | 3 +- tools/binman/etype/ti_secure_rom.py | 3 +- tools/binman/etype/x509_cert.py | 8 +++- tools/binman/test/security/x509_cert.dts | 5 ++- 6 files changed, 55 insertions(+), 49 deletions(-)
Signature · 1 line
2.55.0
PATCH 1/6 binman: openssl: refactor creation of Distinguished Name section from dict
Instead of manually expanding the req_dist_name_dict, create a helper function for creating a config section from any dict. That will also automatically allow other fields to be emitted.
Eventually, we could allow the values to not necesssarily be a single string, but a list of strings, e.g. "OU" -> ["First OU", "Secound OU"], and encode that in the config file using
1.OU = First OU
2.OU = Second OU
as documented in 'man 5 config'.
--- tools/binman/btool/openssl.py | 44 +++++++++++------------------------ 1 file changed, 14 insertions(+), 30 deletions(-)
tools/binman/btool/openssl.py +14 -30
@@ -36,6 +36,14 @@ class Bintoolopenssl(bintool.Bintool): name, 'openssl cryptography toolkit', version_regex=r'OpenSSL (.*) \(', version_args='version') + def dict_to_config_section(self, section_name, d): + if not d: + return '' + s = f'[ {section_name} ]\n' + for field, value in d.items(): + s += f'{field:22s} = {value}\n' + return s + def x509_cert(self, cert_fname, input_fname, key_fname, cn, revision, config_fname): """Create a certificate
@@ -92,8 +100,7 @@ imageSize = INTEGER:{len(indata)} sw_rev (int): Software revision config_fname (str): Filename to write fconfig into req_dist_name_dict (dict): Dictionary containing key-value pairs of - req_distinguished_name section extensions, must contain extensions for - C, ST, L, O, OU, CN and emailAddress + req_distinguished_name section extensions firewall_cert_data (dict): - auth_in_place (int): The Priv ID for copying as the specific host in firewall protected region
@@ -114,14 +121,7 @@ x509_extensions = v3_ca prompt = no dirstring_type = nobmp -[ req_distinguished_name ] -C = {req_dist_name_dict['C']} -ST = {req_dist_name_dict['ST']} -L = {req_dist_name_dict['L']} -O = {req_dist_name_dict['O']} -OU = {req_dist_name_dict['OU']} -CN = {req_dist_name_dict['CN']} -emailAddress = {req_dist_name_dict['emailAddress']} +{self.dict_to_config_section('req_distinguished_name', req_dist_name_dict)} [ v3_ca ] basicConstraints = CA:true
@@ -163,8 +163,7 @@ numFirewallRegions = INTEGER:{firewall_cert_data['num_firewalls']} sw_rev (int): Software revision config_fname (str): Filename to write fconfig into req_dist_name_dict (dict): Dictionary containing key-value pairs of - req_distinguished_name section extensions, must contain extensions for - C, ST, L, O, OU, CN and emailAddress + req_distinguished_name section extensions cert_type (int): Certification type bootcore (int): Booting core bootcore_opts(int): Booting core option, lockstep (0) or split (2) mode
@@ -184,14 +183,7 @@ numFirewallRegions = INTEGER:{firewall_cert_data['num_firewalls']} prompt = no dirstring_type = nobmp - [ req_distinguished_name ] -C = {req_dist_name_dict['C']} -ST = {req_dist_name_dict['ST']} -L = {req_dist_name_dict['L']} -O = {req_dist_name_dict['O']} -OU = {req_dist_name_dict['OU']} -CN = {req_dist_name_dict['CN']} -emailAddress = {req_dist_name_dict['emailAddress']} +{self.dict_to_config_section('req_distinguished_name', req_dist_name_dict)} [ v3_ca ] basicConstraints = CA:true
@@ -252,8 +244,7 @@ emailAddress = {req_dist_name_dict['emailAddress']} sw_rev (int): Software revision config_fname (str): Filename to write fconfig into req_dist_name_dict (dict): Dictionary containing key-value pairs of - req_distinguished_name section extensions, must contain extensions for - C, ST, L, O, OU, CN and emailAddress + req_distinguished_name section extensions cert_type (int): Certification type bootcore (int): Booting core load_addr (int): Load address of image
@@ -274,14 +265,7 @@ x509_extensions = v3_ca prompt = no dirstring_type = nobmp -[ req_distinguished_name ] -C = {req_dist_name_dict['C']} -ST = {req_dist_name_dict['ST']} -L = {req_dist_name_dict['L']} -O = {req_dist_name_dict['O']} -OU = {req_dist_name_dict['OU']} -CN = {req_dist_name_dict['CN']} -emailAddress = {req_dist_name_dict['emailAddress']} +{self.dict_to_config_section('req_distinguished_name', req_dist_name_dict)} [ v3_ca ] basicConstraints = CA:true
Signature · 1 line
2.55.0
Re: [PATCH 1/6] binman: openssl: refactor creation of Distinguished Name section from dict
Hi Rasmus,
Quoted 15 lines · Rasmus Villemoes
binman: openssl: refactor creation of Distinguished Name section from dict
Instead of manually expanding the
req_dist_name_dict, create a helper function for creating a config section from any dict. That will also automatically allow other fields to be emitted.Eventually, we could allow the values to not necesssarily be a single string, but a list of strings, e.g. 'OU' -> ["First OU", "Secound OU"], and encode that in the config file using
1.OU = First OU
2.OU = Second OUas documented in 'man 5 config'.
Quoted 7 lines
Signed-off-by Rasmus Villemoes <[email protected]>
tools/binman/btool/openssl.py| 44 ++++++++++++++-----------------------------1 file changed, 14 insertions(+), 30 deletions(-)Eventually, we could allow the values to not necesssarily be a single string, but a list of strings, e.g. 'OU' -> ["First OU", "Secound OU"],
Two typos: 'necesssarily' and 'Secound'
tools/binman/btool/openssl.py class Bintoolopenssl(bintool.Bintool): · line 36
Quoted 3 lines
tools/binman/btool/openssl.py +1
-0@@ -36,6 +36,14 @@ class Bintoolopenssl(bintool.Bintool): + def dict_to_config_section(self, section_name, d):
Please can you add a docstring with Args and Returns, as with the other methods in this file? Since the method does not use 'self' it could be a @staticmethod, and a more descriptive name than 'd' would help, e.g. data
tools/binman/btool/openssl.py class Bintoolopenssl(bintool.Bintool): · line 37
Quoted 4 lines
tools/binman/btool/openssl.py +2
-0@@ -36,6 +36,14 @@ class Bintoolopenssl(bintool.Bintool): + if not d: + return ''
At this point in the series no caller can pass an empty dict, since ti_secure and ti_secure_rom always set up the full dict, so this branch is dead code here - for me it shows up as uncovered in 'binman test -T', which requires 100% coverage.
Regards,
Simon
Re: [PATCH 1/6] binman: openssl: refactor creation of Distinguished Name section from dict
In reply to Rasmus Villemoestools/binman/btool/openssl.py class Bintoolopenssl(bintool.Bintool): · line 41
Quoted 11 lines · Rasmus Villemoes
tools/binman/btool/openssl.py +3
-0@@ -36,6 +36,14 @@ class Bintoolopenssl(bintool.Bintool): name, 'openssl cryptography toolkit', version_regex=r'OpenSSL (.*) \(', version_args='version') + def dict_to_config_section(self, section_name, d): + if not d: + return ''
Either remove this dead code or perhaps return a bare section header instead of an empty string so that config is at least structurally correct.
Signature · 1 line
Neha Malcom Francis <[email protected]>
PATCH 2/6 binman: x509_cert: allow and parse distinguished-name subnode
In reply to Rasmus VillemoesCurrently, the ti-secure and ti-secure-rom types hardcode the contents of the Distinguished Name section. Those values are not necessarily appropriate for all boards that just happen to include a SOC produced by TI.
Allow placing a distinguished-name subnode inside the binman node representing the Entry_x509_cert (or derived classes), and if present, use that information.
--- tools/binman/etype/ti_secure.py | 3 ++- tools/binman/etype/ti_secure_rom.py | 3 ++- tools/binman/etype/x509_cert.py | 5 +++++ 3 files changed, 9 insertions(+), 2 deletions(-)
tools/binman/etype/ti_secure.py +2 -1
@@ -117,7 +117,8 @@ class Entry_ti_secure(Entry_x509_cert): self.firewall_cert_data['auth_in_place'] = auth_in_place self.ReadFirewallNode() self.sha = fdt_util.GetInt(self._node, 'sha', 512) - self.req_dist_name = {'C': 'US', + if self.req_dist_name is None: + self.req_dist_name = {'C': 'US', 'ST': 'TX', 'L': 'Dallas', 'O': 'Texas Instruments Incorporated',
tools/binman/etype/ti_secure_rom.py +2 -1
@@ -80,7 +80,8 @@ class Entry_ti_secure_rom(Entry_x509_cert): self.dm_data = fdt_util.GetBool(self._node, 'dm-data', False) if self.dm_data: self.load_addr_dm_data = fdt_util.GetInt(self._node, 'load-dm-data', 0x00000000) - self.req_dist_name = {'C': 'US', + if self.req_dist_name is None: + self.req_dist_name = {'C': 'US', 'ST': 'TX', 'L': 'Dallas', 'O': 'Texas Instruments Incorporated',
tools/binman/etype/x509_cert.py +5 -0
@@ -61,6 +61,11 @@ class Entry_x509_cert(Entry_collection): self.key_fname = self.GetEntryArgsOrProps([ EntryArg('keyfile', str)], required=True)[0] self.sw_rev = fdt_util.GetInt(self._node, 'sw-rev', 1) + dist_name_node = self._node.FindNode('distinguished-name') + if dist_name_node: + self.req_dist_name = dict() + for pname, prop in dist_name_node.props.items(): + self.req_dist_name[pname] = prop.value def GetCertificate(self, required, type='generic'): """Get the contents of this entry
Signature · 1 line
2.55.0
Re: [PATCH 2/6] binman: x509_cert: allow and parse distinguished-name subnode
Hi Rasmus,
Quoted 20 lines · Rasmus Villemoes
binman:
x509_cert: allow and parse distinguished-name subnodeCurrently, the ti-secure and ti-secure-rom types hardcode the contents of the Distinguished Name section. Those values are not necessarily appropriate for all boards that just happen to include a SOC produced by TI.
Allow placing a distinguished-name subnode inside the binman node representing the
Entry_x509_cert(or derived classes), and if present, use that information.Signed-off-by Rasmus Villemoes <[email protected]>
tools/binman/etype/ti_secure.py| 3 ++-tools/binman/etype/ti_secure_rom.py | 3 ++- tools/binman/etype/x509_cert.py | 5 +++++ 3 files changed, 9 insertions(+), 2 deletions(-)
tools/binman/etype/x509_cert.py class Entry_x509_cert(Entry_collection): · line 65
Quoted 5 lines
tools/binman/etype/x509_cert.py +5
-0@@ -61,6 +61,11 @@ class Entry_x509_cert(Entry_collection): + dist_name_node = self._node.FindNode('distinguished-name') + if dist_name_node: + self.req_dist_name = dict() + for pname, prop in dist_name_node.props.items(): + self.req_dist_name[pname] = prop.value
Please can you document the new distinguished-name subnode in the class docstring, since the entry documentation is generated from these docstrings?
The ti-secure and ti-secure-rom docstrings should mention it too, along with a note that it overrides the default TI values.
Also, this iterates the raw props, so a property with no value (a boolean) or one with multiple strings ends up formatted as 'True' or as a Python list in the openssl config. I suspect it is worth validating that each value is a single string and raising an error otherwise, at least until the multi-value support mentioned in patch 1 is added. What do you think?
Regards,
Simon
PATCH 3/6 binman: x509: fix CN emitted for basic x509 certificates
In reply to Rasmus VillemoesI'm pretty sure it was never the intention to emit the output filename (cert_fname) as the value of the CN distinguished name.
The cert-ca property was parsed from the dts node, stored in ._cert_ca, passed on as the cn argument to x509_cert(), but then not used at all.
Now that we always read a distinguished-name subnode if present, use that method instead.
--- tools/binman/btool/openssl.py | 7 +++---- tools/binman/etype/x509_cert.py | 3 +-- tools/binman/test/security/x509_cert.dts | 5 ++++- 3 files changed, 8 insertions(+), 7 deletions(-)
tools/binman/btool/openssl.py +3 -4
@@ -44,8 +44,8 @@ class Bintoolopenssl(bintool.Bintool): s += f'{field:22s} = {value}\n' return s - def x509_cert(self, cert_fname, input_fname, key_fname, cn, revision, - config_fname): + def x509_cert(self, cert_fname, input_fname, key_fname, req_dist_name_dict, + revision, config_fname): """Create a certificate Args:
@@ -68,8 +68,7 @@ x509_extensions = v3_ca prompt = no dirstring_type = nobmp -[ req_distinguished_name ] -CN = {cert_fname} +{self.dict_to_config_section('req_distinguished_name', req_dist_name_dict)} [ v3_ca ] basicConstraints = CA:true
tools/binman/etype/x509_cert.py +1 -2
@@ -56,7 +56,6 @@ class Entry_x509_cert(Entry_collection): def ReadNode(self): super().ReadNode() - self._cert_ca = fdt_util.GetString(self._node, 'cert-ca') self._cert_rev = fdt_util.GetInt(self._node, 'cert-revision-int', 0) self.key_fname = self.GetEntryArgsOrProps([ EntryArg('keyfile', str)], required=True)[0]
@@ -96,7 +95,7 @@ class Entry_x509_cert(Entry_collection): cert_fname=output_fname, input_fname=input_fname, key_fname=self.key_fname, - cn=self._cert_ca, + req_dist_name_dict=self.req_dist_name, revision=self._cert_rev, config_fname=config_fname) elif type == 'sysfw':
tools/binman/test/security/x509_cert.dts +4 -1
@@ -8,9 +8,12 @@ binman { x509-cert { - cert-ca = "IOT2050 Firmware Signature"; cert-revision-int = <0>; content = <&u_boot>; + + distinguished-name { + CN = "IOT2050 Firmware Signature"; + }; }; u_boot: u-boot {
Signature · 1 line
2.55.0
Re: [PATCH 3/6] binman: x509: fix CN emitted for basic x509 certificates
Hi Rasmus,
Quoted 21 lines · Rasmus Villemoes
binman: x509: fix CN emitted for basic x509 certificates
I'm pretty sure it was never the intention to emit the output filename (
cert_fname) as the value of the CN distinguished name.The cert-ca property was parsed from the dts node, stored in ._cert_ca, passed on as the cn argument to
x509_cert(), but then not used at all.Now that we always read a distinguished-name subnode if present, use that method instead.
Signed-off-by Rasmus Villemoes <[email protected]>
tools/binman/btool/openssl.py| 7 +++----tools/binman/etype/x509_cert.py | 3 +-- tools/binman/test/security/x509_cert.dts | 5 ++++- 3 files changed, 8 insertions(+), 7 deletions(-)
tools/binman/btool/openssl.py x509_extensions = v3_ca · line 68
Quoted 3 lines
tools/binman/btool/openssl.py +1
-2@@ -68,8 +68,7 @@ x509_extensions = v3_ca -[ req_distinguished_name ] -CN = {cert_fname} +{self.dict_to_config_section('req_distinguished_name', req_dist_name_dict)}
If the entry has no distinguished-name subnode, req_dist_name_dict is None here (the ti-secure types fall back to their hardcoded defaults, but the plain x509-cert type has no fallback). dict_to_config_section() then returns an empty string, so the config still says 'distinguished_name = req_distinguished_name' but the section does not exist. I tried this and openssl fails with:
Unable to get 'req_distinguished_name' section Error making certificate request
So this patch silently makes the subnode mandatory for the generic type, with a rather cryptic build failure for anyone using x509-cert without it (previously it worked, emitting the filename as CN). Please can you either keep a fallback (e.g. CN = cert_fname as before), or raise a proper error from the etype when the subnode is missing, so the user gets a helpful message pointing at the node? A test for the missing-subnode case would cover the 'if not d' branch added in patch 1 too.
tools/binman/btool/openssl.py class Bintoolopenssl(bintool.Bintool): · line 45
Quoted 6 lines
tools/binman/btool/openssl.py +2
-2@@ -44,8 +44,8 @@ class Bintoolopenssl(bintool.Bintool): - def x509_cert(self, cert_fname, input_fname, key_fname, cn, revision, - config_fname): + def x509_cert(self, cert_fname, input_fname, key_fname, req_dist_name_dict, + revision, config_fname):
The Args docstring below still documents 'cn (str): Common name' but the parameter is now req_dist_name_dict, so please can you update it to match, as done for the other functions in patch 1?
Regards,
Simon
PATCH 4/6 k3-binman.dtsi: add keyfile_template node
In reply to Rasmus VillemoesCurrently, it is somewhat tedious for downstream/custom boards to override the keyfile used for signing all the various artifacts involved in booting on K3 devices to point at their own key (or pkcs#11 uri for that matter). One has to have overrides in one's .dtsi for each and every ti-secure / ti-secure-rom node, and a lot of those do not even have dts labels, so one has to use their full /binman/... path.
Add a small dedicated template that only contains the keyfile property. In a subsequent patch, we will change nodes that contain an explicit 'keyfile = "custMpk.pem"' to instead have this template inserted. This way, the downstream board's .dtsi file only needs to contain something like
&keyfile_template {
keyfile = "/path/to/own/key.pem";
};and that will then automatically propagate to all the nodes.
--- arch/arm/dts/k3-binman.dtsi | 4 ++++ 1 file changed, 4 insertions(+)
arch/arm/dts/k3-binman.dtsi +4 -0
@@ -25,6 +25,10 @@ filename = "arch/arm/mach-k3/keys/ti-degenerate-key.pem"; }; }; + + keyfile_template: template-keyfile { + keyfile = "custMpk.pem"; + }; }; #ifndef CONFIG_ARM64
Signature · 1 line
2.55.0
Re: [PATCH 4/6] k3-binman.dtsi: add keyfile_template node
Hi Rasmus,
Quoted 27 lines · Rasmus Villemoes
k3-binman.dtsi: addkeyfile_templatenodeCurrently, it is somewhat tedious for downstream/custom boards to override the keyfile used for signing all the various artifacts involved in booting on K3 devices to point at their own key (or pkcs#11 uri for that matter). One has to have overrides in one's .dtsi for each and every ti-secure / ti-secure-rom node, and a lot of those do not even have dts labels, so one has to use their full /binman/... path.
Add a small dedicated template that only contains the keyfile property. In a subsequent patch, we will change nodes that contain an explicit 'keyfile = 'custMpk.pem'' to instead have this template inserted. This way, the downstream board's .dtsi file only needs to contain something like
&keyfile_template { keyfile = '/path/to/own/key.pem'; };[...]
arch/arm/dts/k3-binman.dtsi| 4 ++++1 file changed, 4 insertions(+)
arch/arm/dts/k3-binman.dtsi line 28
Quoted 4 lines
arch/arm/dts/k3-binman.dtsi +4
-0@@ -25,6 +25,10 @@ + + keyfile_template: template-keyfile { + keyfile = 'custMpk.pem'; + };
This looks good: since merge_props() only copies properties missing from the target node, an explicit keyfile in a ti-secure node still takes precedence, as your cover letter says.
Since the whole point is discoverability, please can you also mention this in doc/board/ti/k3.rst, which already discusses signing with custMpk? A short note showing the
&keyfile_template {
keyfile = "/path/to/own/key.pem";
};override would help people find it without reading the dtsi.
Regards,
Simon
PATCH 5/6 k3-binman.dtsi: add empty distinguished_name_template
In reply to Rasmus VillemoesWith the previous commit, "binman: x509_cert: allow and parse distinguished-name subnode", it is possible for downstream boards to put more appropriate values in the req_distinguished_name section in the certificate being generated than the hard-coded TI values. However, that requires adding a distinguished-name section to each and every ti-secure or ti-secure-rom binman node.
In order to avoid lots of repetition and make it easier to just add the information in one place, add an empty template-dist-name template which we can inject in every such binman node. That makes no difference for existing boards, but makes it possible for downstream boards to simply do
&distinguished_name_template {
distinguished-name {
C = "DK";
...
};
};in their -u-boot.dtsi.
--- arch/arm/dts/k3-binman.dtsi | 10 ++++++++++ 1 file changed, 10 insertions(+)
arch/arm/dts/k3-binman.dtsi +10 -0
@@ -29,6 +29,16 @@ keyfile_template: template-keyfile { keyfile = "custMpk.pem"; }; + + distinguished_name_template: template-dist-name { + /* + * Empty, but allows downstreams to inject a + * distinguished-name subnode with appropriate + * contents from their -u-boot.dtsi, which will + * propagate to all ti-secure(-rom) nodes including + * this template. + */ + }; }; #ifndef CONFIG_ARM64
Signature · 1 line
2.55.0
Re: [PATCH 5/6] k3-binman.dtsi: add empty distinguished_name_template
Quoted 22 lines · Rasmus Villemoes
k3-binman.dtsi: add emptydistinguished_name_templateWith the previous commit, "binman:
x509_cert: allow and parse distinguished-name subnode", it is possible for downstream boards to put more appropriate values in thereq_distinguished_namesection in the certificate being generated than the hard-coded TI values. However, that requires adding a distinguished-name section to each and every ti-secure or ti-secure-rom binman node.In order to avoid lots of repetition and make it easier to just add the information in one place, add an empty template-dist-name template which we can inject in every such binman node. That makes no difference for existing boards, but makes it possible for downstream boards to simply do
&distinguished_name_template { distinguished-name { C = 'DK'; ... }; };
Quoted 6 lines
in their
-u-boot.dtsi.Signed-off-by Rasmus Villemoes <[email protected]>
arch/arm/dts/k3-binman.dtsi| 10 ++++++++++1 file changed, 10 insertions(+)
PATCH 6/6 k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodes
In reply to Rasmus VillemoesThis should not make any functional difference, and downstream boards that already override the keyfile property via an explicit .dtsi fragment for each node will continue to have precedence over the value coming from the template. But this makes it much more ergonomic to have the keyfile and distinguished-name have consistent values throughout, because one only has to override the values in the templates and not each and every individual node.
--- arch/arm/dts/k3-binman.dtsi | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)
arch/arm/dts/k3-binman.dtsi +10 -10
@@ -150,8 +150,8 @@ arch = "arm"; compression = "none"; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&board_cfg>; - keyfile = "custMpk.pem"; }; board_cfg: board-cfg { filename = "board-cfg.bin";
@@ -165,8 +165,8 @@ arch = "arm"; compression = "none"; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&pm_cfg>; - keyfile = "custMpk.pem"; }; pm_cfg: pm-cfg { filename = "pm-cfg.bin";
@@ -179,8 +179,8 @@ arch = "arm"; compression = "none"; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&rm_cfg>; - keyfile = "custMpk.pem"; }; rm_cfg: rm-cfg { filename = "rm-cfg.bin";
@@ -193,8 +193,8 @@ arch = "arm"; compression = "none"; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&sec_cfg>; - keyfile = "custMpk.pem"; }; sec_cfg: sec-cfg { filename = "sec-cfg.bin";
@@ -287,8 +287,8 @@ load = <CONFIG_K3_ATF_LOAD_ADDR>; entry = <CONFIG_K3_ATF_LOAD_ADDR>; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&atf>; - keyfile = "custMpk.pem"; }; atf: atf-bl31 { };
@@ -303,8 +303,8 @@ load = <CONFIG_K3_OPTEE_LOAD_ADDR>; entry = <CONFIG_K3_OPTEE_LOAD_ADDR>; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&tee>; - keyfile = "custMpk.pem"; }; tee: tee-os { optional;
@@ -330,8 +330,8 @@ load = <CONFIG_SPL_TEXT_BASE>; entry = <CONFIG_SPL_TEXT_BASE>; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&u_boot_spl_nodtb>; - keyfile = "custMpk.pem"; }; u_boot_spl_nodtb: blob-ext {
@@ -419,8 +419,8 @@ compression = "none"; load = <CONFIG_TEXT_BASE>; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&u_boot_nodtb>; - keyfile = "custMpk.pem"; }; u_boot_nodtb: u-boot-nodtb { };
@@ -521,8 +521,8 @@ load = <CONFIG_K3_ATF_LOAD_ADDR>; entry = <CONFIG_K3_ATF_LOAD_ADDR>; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&atf_falcon>; - keyfile = "custMpk.pem"; }; atf_falcon: atf-bl31 { };
@@ -536,8 +536,8 @@ load = <CONFIG_K3_OPTEE_LOAD_ADDR>; entry = <CONFIG_K3_OPTEE_LOAD_ADDR>; ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&tee_falcon>; - keyfile = "custMpk.pem"; }; tee_falcon: tee-os { optional;
Signature · 1 line
2.55.0
Re: [PATCH 6/6] k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodes
Hi Rasmus,
Quoted 10 lines · Rasmus Villemoes
k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodesThis should not make any functional difference, and downstream boards that already override the keyfile property via an explicit .dtsi fragment for each node will continue to have precedence over the value coming from the template. But this makes it much more ergonomic to have the keyfile and distinguished-name have consistent values throughout, because one only has to override the values in the templates and not each and every individual node.
Quoted 6 lines
Signed-off-by Rasmus Villemoes <[email protected]>
arch/arm/dts/k3-binman.dtsi| 20 ++++++++++----------1 file changed, 10 insertions(+), 10 deletions(-)
k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodes
I still see lots of custMpk.pem in .dtsi files after this series so I'm not sure of the exact mechanism for your claim in this series. I was expecting it to remove all of those?
arch/arm/dts/k3-binman.dtsi line 334
Quoted 8 lines
arch/arm/dts/k3-binman.dtsi +1
-1@@ -330,8 +330,8 @@ ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&u_boot_spl_nodtb>; - keyfile = 'custMpk.pem'; };
Removing the keyfile property leaves a stray blank line before the closing brace. Please can you drop that while you are at it?
Regards,
Simon
Re: [PATCH 6/6] k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodes
Quoted 15 lines · Simon Glass
Hi Rasmus,
Quoted 12 lines · Rasmus Villemoes
k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodesThis should not make any functional difference, and downstream boards that already override the keyfile property via an explicit .dtsi fragment for each node will continue to have precedence over the value coming from the template. But this makes it much more ergonomic to have the keyfile and distinguished-name have consistent values throughout, because one only has to override the values in the templates and not each and every individual node.
Signed-off-by Rasmus Villemoes <[email protected]>
Quoted 8 lines
Quoted 4 lines
arch/arm/dts/k3-binman.dtsi| 20 ++++++++++----------1 file changed, 10 insertions(+), 10 deletions(-)
k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodesI still see lots of custMpk.pem in .dtsi files after this series so I'm not sure of the exact mechanism for your claim in this series. I was expecting it to remove all of those?
Well, changing those are not needed for my immediate purpose, I only need those in k3-binman.dtsi to be updated, as that is the only upstream file I include. Yes, they should probably all be updated, but I didn't want to pollute this series with tons of mechanical changes, that can be done as a followup if and when this lands.
Quoted 11 lines
arch/arm/dts/k3-binman.dtsi line 334
Quoted 8 lines
arch/arm/dts/k3-binman.dtsi +1
-1@@ -330,8 +330,8 @@ ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&u_boot_spl_nodtb>; - keyfile = 'custMpk.pem'; };Removing the keyfile property leaves a stray blank line before the closing brace. Please can you drop that while you are at it?
Of course.
Rasmus
Re: [PATCH 6/6] k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodes
Hi Rasmus,
Quoted 27 lines · Rasmus Villemoes
Quoted 15 lines · Simon Glass
Hi Rasmus,
Quoted 12 lines · Rasmus Villemoes
k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodesThis should not make any functional difference, and downstream boards that already override the keyfile property via an explicit .dtsi fragment for each node will continue to have precedence over the value coming from the template. But this makes it much more ergonomic to have the keyfile and distinguished-name have consistent values throughout, because one only has to override the values in the templates and not each and every individual node.
Signed-off-by Rasmus Villemoes <[email protected]>Quoted 8 lines
Quoted 4 lines
arch/arm/dts/k3-binman.dtsi| 20 ++++++++++----------1 file changed, 10 insertions(+), 10 deletions(-)
k3-binman.dtsi: insert keyfile and distinguished name templates in all ti-secure(-rom) nodesI still see lots of custMpk.pem in .dtsi files after this series so I'm not sure of the exact mechanism for your claim in this series. I was expecting it to remove all of those?
Quoted 5 lines
Well, changing those are not needed for my immediate purpose, I only need those in
k3-binman.dtsito be updated, as that is the only upstream file I include. Yes, they should probably all be updated, but I didn't want to pollute this series with tons of mechanical changes, that can be done as a followup if and when this lands.
Fair enough, thanks.
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
arch/arm/dts/k3-binman.dtsi +1
-1@@ -330,8 +330,8 @@ ti-secure { + insert-template = <&keyfile_template>, <&distinguished_name_template>; content = <&u_boot_spl_nodtb>; - keyfile = 'custMpk.pem';
Quoted 7 lines
Quoted 4 lines
arch/arm/dts/k3-binman.dtsi line 334
Quoted 1 line
arch/arm/dts/k3-binman.dtsi +0
-0};Removing the keyfile property leaves a stray blank line before the closing brace. Please can you drop that while you are at it?
Of course.
Regards,
Simon
Quoted 1 line
Rasmus
Quoted 7 lines · Rasmus Villemoes
When trying to hook up use of a customer's own signing infrastructure for some am62x devices, I found it was more difficult than it should be to point at the right key(s). One has to patch in the keyfile= property in way too many places, and if you miss just one, the build succeeds using the generic custMpk.pem for the one you missed, but of course the resulting artifacts are unbootable.
The default value for keyfile i.e custMpk.pem is generated via binman from the custMpk node in binman (defined in k3-binman.dtsi), modifying the filepath to your custom key should work as a single line change as is.
Perhaps this could be made more clear by the use of a macro instead of hardcoding the string "custMpk.pem" in all the places to make it more obvious.
Other than that, the addition of configurable distinguished-name could be valuable addition, do you have any specific use case asides from downstream boards? I wonder if it could be valuable for existing upstream platforms from Toradex/Phytec.
CC+: Francesco, Wadim
Quoted 33 lines
Use the template mechanism that binman already has to allow one to only set the keyfile= in a single place. I cannot really imagine any case where one would not want to use the same key for signing all artifacts across the r5/a53 cores, but that is of course still possible, just as any downstreams that have explicit
&{/binman/tiboot3-am62x-hs-evm.bin/ti-secure-rom} { keyfile = "foobar.pem"; };&{/binman/tiboot3-am62x-hs-fs-evm.bin/ti-secure-rom} { keyfile = "foobar.pem"; };will continue to work.
Rasmus Villemoes (6):
binman: openssl: refactor creation of Distinguished Name section from
dict
binman:x509_cert: allow and parse distinguished-name subnode
binman: x509: fix CN emitted for basic x509 certificatesk3-binman.dtsi: addkeyfile_templatenode
k3-binman.dtsi: add emptydistinguished_name_templatek3-binman.dtsi: insert keyfile and distinguished name templates in all
ti-secure(-rom) nodesarch/arm/dts/k3-binman.dtsi | 34 +++++++++++----- tools/binman/btool/openssl.py | 51 ++++++++---------------- tools/binman/etype/ti_secure.py | 3 +- tools/binman/etype/ti_secure_rom.py | 3 +- tools/binman/etype/x509_cert.py | 8 +++- tools/binman/test/security/x509_cert.dts | 5 ++- 6 files changed, 55 insertions(+), 49 deletions(-)
Quoted 9 lines · Anshul Dalal
Quoted 7 lines · Rasmus Villemoes
When trying to hook up use of a customer's own signing infrastructure for some am62x devices, I found it was more difficult than it should be to point at the right key(s). One has to patch in the keyfile= property in way too many places, and if you miss just one, the build succeeds using the generic custMpk.pem for the one you missed, but of course the resulting artifacts are unbootable.
Quoted 5 lines
The default value for keyfile i.e custMpk.pem is generated via binman from the custMpk node in binman (defined in
k3-binman.dtsi), modifying the filepath to your custom key should work as a single line change as is.
No, because I don't have and will not have the "key" as an actual file. And nobody who really cares about security should expose their signing keys to the build environment. So I cannot simply change how the custMpk.pem key in the (top-level) build-directory gets generated from the one in the source tree.
I'm using a pkcs11:object=... uri as the keyfile property - that works just fine out-of-the-box with a suitable OPENSSL_CONF in the environment, so that openssl picks up the right pkcs11 provider module and gets a signature that way.
So I really need a way for having a single place to set keyfile = "pkcs11:object=..." and have that be used in all relevant nodes. Yes, I see that I should have emphasized that use case more in the commit log for patch 4, using that as example instead of "/path/to/own/key.pem".
[In my concrete case, the signing is done by a remote service, but it would really be the same issue if one were using a yubikey or some other local HSM].
Quoted 3 lines
Perhaps this could be made more clear by the use of a macro instead of hardcoding the string "custMpk.pem" in all the places to make it more obvious.
Sure, we could introduce a macro and do
#ifndef K3_KEYFILE
#define K3_KEYFILE "custMpk.pem"
#endif
at the top of k3-binman.dtsi, but since binman already has the template mechanism, that seemed a lot more appropriate to use.
Rasmus
Quoted 28 lines · Rasmus Villemoes
Quoted 9 lines · Anshul Dalal
Quoted 7 lines · Rasmus Villemoes
When trying to hook up use of a customer's own signing infrastructure for some am62x devices, I found it was more difficult than it should be to point at the right key(s). One has to patch in the keyfile= property in way too many places, and if you miss just one, the build succeeds using the generic custMpk.pem for the one you missed, but of course the resulting artifacts are unbootable.
Quoted 5 lines
The default value for keyfile i.e custMpk.pem is generated via binman from the custMpk node in binman (defined in
k3-binman.dtsi), modifying the filepath to your custom key should work as a single line change as is.No, because I don't have and will not have the "key" as an actual file. And nobody who really cares about security should expose their signing keys to the build environment. So I cannot simply change how the custMpk.pem key in the (top-level) build-directory gets generated from the one in the source tree.
I'm using a pkcs11:object=... uri as the keyfile property - that works just fine out-of-the-box with a suitable
OPENSSL_CONFin the environment, so that openssl picks up the right pkcs11 provider module and gets a signature that way.
Quoted 9 lines
So I really need a way for having a single place to set keyfile = "pkcs11:object=..." and have that be used in all relevant nodes. Yes, I see that I should have emphasized that use case more in the commit log for patch 4, using that as example instead of "/path/to/own/key.pem".
[In my concrete case, the signing is done by a remote service, but it would really be the same issue if one were using a yubikey or some other local HSM].
I see the use case now, what we have here makes sense. To add to what Simon said[1] a small section documenting the use of pkcss11:object inside the node in the k3.rst doc would be good to see.
[1]: https://lore.kernel.org/u-boot/CAFLszTh=4fWg6L74vnaJz7O9kOxE68XP3BCJEWVnBvc1aopHJg@mail.gmail.com/
Quoted 4 lines
Quoted 3 lines
Perhaps this could be made more clear by the use of a macro instead of hardcoding the string "custMpk.pem" in all the places to make it more obvious.
Quoted 8 lines
Sure, we could introduce a macro and do
#ifndef
K3_KEYFILE
#defineK3_KEYFILE"custMpk.pem"
#endifat the top of
k3-binman.dtsi, but since binman already has the template mechanism, that seemed a lot more appropriate to use.
That's fair, we wouldn't need a macro if we replace all instances of custMpk with the template. I can take that up once this is merged.
Thanks for the patch,
Anshul
Hi Rasmus,
Quoted 2 lines · Rasmus Villemoes
Use the template mechanism that binman already has to allow one to only set the keyfile= in a single place.
Thanks for sorting this out - having to override the keyfile in every node is definitely error-prone. I think this is a nice use of templates. Do we need to explicitly document how to specify the key, etc.? I don't see doc/ changes in this series.
Regards,
Simon