boot: fit: authenticate the dm-verity roothash
v4 cover boot: fit: authenticate the dm-verity roothash
A signed FIT configuration can delegate the integrity of a (potentially large) root filesystem image to the kernel's dm-verity instead of having U-Boot hash the whole payload at boot: the FIT carries a "dm-verity" subnode with the roothash, salt and block parameters, U-Boot passes the roothash to Linux through the dm-mod.create bootargs, and dm-verity then validates the filesystem block by block against it.
For that to be safe the roothash has to be trusted, and in a signed configuration the only thing that establishes trust is the configuration signature. The roothash was not covered by it. fit_config_add_hash() collected the image node, its hash subnodes and its cipher subnode into the signed region, but not the dm-verity subnode, so the roothash, the sole integrity anchor for the filesystem, was left unsigned.
The result is a verified-boot bypass for the root filesystem: an attacker who can rewrite the boot medium can replace the filesystem, recompute a matching dm-verity tree, write the new roothash into the unsigned dm-verity subnode, and the configuration signature still verifies. dm-verity then faithfully validates the malicious filesystem against the attacker's roothash.
This series closes the gap.
v4: fix the CI failure Tom Rini reported
* test_fit_verity_roothash_signed() built a FIT with a dm-verity subnode, which makes mkimage shell out to veritysetup internally to compute the Merkle tree/roothash, but the test only declared dtc, fdtget, fdtput and openssl as required tools. On the CI image, which doesn't have veritysetup installed, mkimage's internal exec of it failed silently (the diagnostic goes to a private pipe mkimage uses to parse veritysetup's output, not to stderr) and the test hard failed with "Can't add verification data ... (Input/output error)" instead of skipping, the way test_fit_verity.py's veritysetup-using tests already do via the same marker.
* add @pytest.mark.requiredtool('veritysetup') to test_fit_verity_roothash_signed(), matching test_fit_verity.py
* collect Reviewed-by tags from Simon Glass on all three patches. Simon also asked, on this cover letter, whether the series should spell out the compatibility break explicitly (an image with a dm-verity subnode signed by an older mkimage stops verifying, and vice versa); resolved in-thread since there are no real-world deployments of this mechanism yet, only the upcoming OpenWrt bootstd use case -- no documentation change needed for this revision
v3: address comments by Simon Glass
* factor tools/image-host.c's node-path collection into patch 1 too, mirroring the boot-side helper, so both the sign side and the verify side share the same shape and stay easy to compare
* use present tense for the pre-patch code description in patch 2; document the dm-verity subnode in the rebuilt node list in doc/usage/fit/signature.rst and note the signature coverage in doc/usage/fit/dm-verity.rst
* in the unit test, bound-check fdt_find_regions()'s returned count the same way fit_config_check_sig() does, tamper the digest through the buffer instead of casting away const, and note in a comment that the digest check stands in for the whole dm-verity node
* fix test_fit_verity_roothash_signed(), which never actually ran: its ITS baked the tmpdir prefix into the /incbin/() paths while dtc also resolves incbin paths relative to the .its file's own directory, so the path was searched doubled and mkimage always failed
* kept Reviewed-by tags on patches 1 and 2: neither change invalidates what was reviewed, the diffs are mechanical/cosmetic
v2: address comments by Tom Rini
* drop the VISIBLE_IF_UT visibility macro; fit_config_get_signed_nodes() is now simply non-static (previously the function would end up being inlined, so there *is* a real cost to this)
* document test_fit_verity_sign.py with pydoc docstrings including an ITS example, and add a page under doc/develop/pytest/ so the module is rendered in the generated documentation
* collect Reviewed-by tags on patches 1 and 2
Daniel Golle (3):
boot: fit: factor out node-path collection in fit_config_add_hash()
boot: fit: cover the dm-verity roothash with the config signature
test: fit: verify dm-verity roothash is covered by the config
signature
boot/image-fit-sig.c | 108 +++++++---- doc/develop/pytest/test_fit_verity_sign.rst | 10 + doc/usage/fit/dm-verity.rst | 5 + doc/usage/fit/signature.rst | 2 +- include/image.h | 23 +++ test/boot/fit_verity.c | 200 +++++++++++++++++++ test/py/tests/test_fit_verity_sign.py | 203 ++++++++++++++++++++ tools/image-host.c | 94 ++++++--- 8 files changed, 579 insertions(+), 66 deletions(-) create mode 100644 doc/develop/pytest/test_fit_verity_sign.rst create mode 100644 test/py/tests/test_fit_verity_sign.py
base-commit: b635d43bca429500cb8ef20aa151cb5773b9a8a5
Signature · 1 line
2.55.0
PATCH v4 1/3 boot: fit: factor out node-path collection in fit_config_add_hash()
Both the boot-side and host-side fit_config_add_hash() repeat the same sequence to append a node's path to the hashed-node list three times:
for the image node, for each hash subnode and for the cipher subnode. Extract it into a helper, fit_config_add_node(), in each file, with no functional change.
---
v4: collect Reviewed-by from Simon Glass
v3: also factor out tools/image-host.c's fit_config_add_hash(), which
had the same duplication, per Simon Glass's review; this keeps the sign-side and verify-side node list construction easy to compare. Kept Tom's Reviewed-by: the boot-side code he reviewed is unchanged and the host-side addition mechanically mirrors the same pattern.
v2: no changes
boot/image-fit-sig.c | 73 ++++++++++++++++++++++++++------------------ tools/image-host.c | 73 ++++++++++++++++++++++++++++---------------- 2 files changed, 91 insertions(+), 55 deletions(-)
boot/image-fit-sig.c +44 -29
@@ -230,6 +230,37 @@ int fit_image_verify_required_sigs(const void *fit, int image_noffset,return 0 ; } +/** +* fit_config_add_node() - Append one node's path to the hashed-node list +* +* @fit: FIT blob +* @noffset: Offset of the node whose path should be added +* @node_inc: Array of path pointers to fill +* @count: Pointer to current count (updated on return) +* @max_nodes: Maximum entries in @node_inc +* @buf: Buffer for packed path strings +* @buf_used: Pointer to bytes used in @buf (updated on return) +* @buf_len: Total size of @buf +* Return: 0 on success, -ve on error +*/ +static int fit_config_add_node (const void *fit,int noffset,char **node_inc, +int *count,int max_nodes,char *buf, +int *buf_used,int buf_len) +{ +int ret, len; + +if (*count >= max_nodes) +return -ENOSPC ; + ret =fdt_get_path (fit, noffset, buf + *buf_used, buf_len - *buf_used); +if (ret <0 ) +return -ENOENT ; + len =strlen (buf + *buf_used) +1 ; + node_inc[(*count)++] = buf + *buf_used; + *buf_used += len; + +return 0 ; +} +/** * fit_config_add_hash() - Add hash nodes for one image to the node list *
@@ -250,18 +281,12 @@ static int fit_config_add_hash(const void *fit, int image_noffset, char **node_inc, int *count, int max_nodes, char *buf, int *buf_used, int buf_len) { - int noffset, hash_count, ret, len; + int noffset, hash_count, ret; -if (*count >= max_nodes) -return -ENOSPC ; - - ret =fdt_get_path (fit, image_noffset, buf + *buf_used, - buf_len - *buf_used); -if (ret <0 ) -return -ENOENT ; - len =strlen (buf + *buf_used) +1 ; - node_inc[(*count)++] = buf + *buf_used; - *buf_used += len; + ret =fit_config_add_node (fit, image_noffset, node_inc, count, + max_nodes, buf, buf_used, buf_len); +if (ret) +return ret;/* Add all this image's hash subnodes */ hash_count =0 ;
@@ -273,15 +298,10 @@ static int fit_config_add_hash(const void *fit, int image_noffset,if (strncmp (name,FIT_HASH_NODENAME ,strlen (FIT_HASH_NODENAME )))continue ; -if (*count >= max_nodes) -return -ENOSPC ; - ret =fdt_get_path (fit, noffset, buf + *buf_used, - buf_len - *buf_used); -if (ret <0 ) -return -ENOENT ; - len =strlen (buf + *buf_used) +1 ; - node_inc[(*count)++] = buf + *buf_used; - *buf_used += len; + ret =fit_config_add_node (fit, noffset, node_inc, count, + max_nodes, buf, buf_used, buf_len); +if (ret) +return ret; hash_count++; }
@@ -296,15 +316,10 @@ static int fit_config_add_hash(const void *fit, int image_noffset,if (noffset != -FDT_ERR_NOTFOUND ) {if (noffset <0 )return -EIO ; -if (*count >= max_nodes) -return -ENOSPC ; - ret =fdt_get_path (fit, noffset, buf + *buf_used, - buf_len - *buf_used); -if (ret <0 ) -return -ENOENT ; - len =strlen (buf + *buf_used) +1 ; - node_inc[(*count)++] = buf + *buf_used; - *buf_used += len; + ret =fit_config_add_node (fit, noffset, node_inc, count, + max_nodes, buf, buf_used, buf_len); +if (ret) +return ret; }return 0 ;
tools/image-host.c +47 -26
104 lines not on this page · raw message
Signature · 1 line
2.55.0
PATCH v4 2/3 boot: fit: cover the dm-verity roothash with the config signature
In reply to Daniel GolleA dm-verity protected filesystem image is not hashed by U-Boot when it is loaded; its integrity is delegated to the kernel, which validates the filesystem on the fly against the roothash taken from the FIT dm-verity subnode. The roothash is therefore the sole integrity anchor for the filesystem, yet fit_config_add_hash() only adds the image node, its hash subnodes and its cipher subnode to the signed region, leaving the dm-verity subnode (roothash, salt and block parameters) unsigned.
An attacker able to rewrite the boot medium could then replace both the filesystem and the roothash, recompute a matching dm-verity tree and keep the configuration signature valid, defeating verified boot for the root filesystem.
Add the dm-verity subnode to the list of nodes covered by the configuration signature, both when signing (tools/image-host.c) and when verifying (boot/image-fit-sig.c), so the roothash and salt are authenticated together with the rest of the configuration.
---
v4: collect Reviewed-by from Simon Glass
v3: use present tense for the pre-patch code description, per Simon
Glass's review; document the dm-verity subnode in the rebuilt node list in doc/usage/fit/signature.rst and note the signature coverage in doc/usage/fit/dm-verity.rst. The tools/image-host.c hunk now calls the patch 1 helper instead of duplicating the pattern inline (no functional change). Kept Tom's Reviewed-by: the functional change he reviewed is unchanged; the rest is doc/message wording and a mechanical refactor.
v2: no changes
boot/image-fit-sig.c | 23 +++++++++++++++++++---- doc/usage/fit/dm-verity.rst | 5 +++++ doc/usage/fit/signature.rst | 2 +- tools/image-host.c | 21 +++++++++++++++++++++ 4 files changed, 46 insertions(+), 5 deletions(-)
boot/image-fit-sig.c +19 -4
@@ -264,8 +264,8 @@ static int fit_config_add_node(const void *fit, int noffset, char **node_inc,/** * fit_config_add_hash() - Add hash nodes for one image to the node list * -* Adds the image path, all its hash-* subnode paths, and its cipher -* subnode path (if present) to the packed buffer. +* Adds the image path, all its hash-* subnode paths, and its cipher and +* dm-verity subnode paths (each if present) to the packed buffer. * * @fit: FIT blob * @image_noffset: Image node offset (e.g. /images/kernel-1)
@@ -322,6 +322,21 @@ static int fit_config_add_hash(const void *fit, int image_noffset,return ret; } +/* +* Add this image's dm-verity node if present. Its roothash is the +* only integrity anchor for a dm-verity filesystem image, so it must +* be covered by the configuration signature. +*/ + noffset =fdt_subnode_offset (fit, image_noffset,FIT_VERITY_NODENAME ); +if (noffset != -FDT_ERR_NOTFOUND ) { +if (noffset <0 ) +return -EIO ; + ret =fit_config_add_node (fit, noffset, node_inc, count, + max_nodes, buf, buf_used, buf_len); +if (ret) +return ret; + } +return 0 ; }
@@ -329,8 +344,8 @@ static int fit_config_add_hash(const void *fit, int image_noffset,* fit_config_get_hash_list() - Build the list of nodes to hash * * Works through every image referenced by the configuration and collects the -* node paths: root + config + all referenced images with their hash and -* cipher subnodes. +* node paths: root + config + all referenced images with their hash, +* cipher and dm-verity subnodes. * * Properties known not to be image references (description, compatible, * default, load-only) are skipped, so any new image type is covered by default.
doc/usage/fit/dm-verity.rst +5 -0
@@ -209,6 +209,11 @@ typically be obtained from its output. The ``digest`` and ``salt`` byte arrays correspond to the hex-encoded ``Root hash`` and ``Salt`` printed by ``veritysetup format``. +When the configuration is signed, ``digest`` and ``salt`` are covered by +the configuration signature (see :doc:`signature`), so the roothash +cannot be swapped out for a matching one without invalidating the +signature. + Optional boolean properties (when present, they are collected and appended as dm-verity optional parameters with hyphens converted to underscores):
doc/usage/fit/signature.rst +1 -1
@@ -359,7 +359,7 @@ however, U-Boot does not read 'hashed-nodes'. Instead it rebuilds the node list from the configuration's own image references (kernel, fdt, ramdisk, etc.), since 'hashed-nodes' is not itself covered by the signature. The rebuilt list always includes the root node, the configuration node, each -referenced image node and its hash/cipher subnodes. +referenced image node and its hash, cipher and dm-verity subnodes. The image is walked in order and each tag processed as follows:
tools/image-host.c +21 -0
@@ -1287,6 +1287,27 @@ static int fit_config_add_hash(const void *fit, int image_noffset,return ret; } +/* +* Add this image's dm-verity node if present. Its roothash is the +* only integrity anchor for a dm-verity filesystem image, so it must +* be covered by the configuration signature. +*/ + noffset =fdt_subnode_offset (fit, image_noffset, +FIT_VERITY_NODENAME ); +if (noffset != -FDT_ERR_NOTFOUND ) { +if (noffset <0 ) { +fprintf (stderr, +"Failed to get dm-verity node in configuration '%s/%s' image '%s': %s\n" , + conf_name, sig_name, iname, +fdt_strerror (noffset)); +return -EIO ; + } + ret =fit_config_add_node (fit, noffset, node_inc, conf_name, + sig_name, iname); +if (ret) +return ret; + } +return 0 ; }
Signature · 1 line
2.55.0
PATCH v4 3/3 test: fit: verify dm-verity roothash is covered by the config signature
In reply to Daniel GolleA dm-verity protected filesystem image is not hashed by U-Boot; its integrity is delegated to the kernel, which trusts the roothash taken from the FIT dm-verity subnode. For that chain of trust to hold, the roothash (and salt) must be part of the region covered by the configuration signature, otherwise an attacker can replace both the filesystem and the roothash while keeping the signature valid.
Add two independent checks of this property:
- test/py/tests/test_fit_verity_sign.py signs a configuration that references a filesystem image carrying a dm-verity subnode, then confirms that tampering the roothash or the salt is rejected by fit_check_sign. A control that tampers a byte known to be signed proves the check can fail. A matching page is added under doc/develop/pytest/ so the module documentation is rendered with the rest of the generated docs.
- test/boot/fit_verity.c gains a runtime unit test that builds the exact node list the configuration signature is computed over, turns it into hashed regions and checks both that the roothash bytes fall inside a signed region and that tampering them changes the hash. It needs no private key, so it also runs on real devices and uses the same hash path a device would.
To let the unit test build the signed-region node list, rename the config node-list helper to fit_config_get_signed_nodes(), make it non-static and declare it in image.h.
---
v4: fix the CI failure Tom Rini reported
* test_fit_verity_roothash_signed() builds a FIT with a dm-verity subnode, which makes mkimage shell out to veritysetup internally to compute the Merkle tree/roothash, but the test only declared dtc, fdtget, fdtput and openssl as required tools. On the CI image, which doesn't have veritysetup installed, mkimage's internal exec of it fails silently (the diagnostic goes to a private pipe mkimage uses to parse veritysetup's output, not to stderr) and the test hard fails with "Can't add verification data ... (Input/output error)" instead of skipping, the way test_fit_verity.py's veritysetup-using tests already do via the same marker.
* add @pytest.mark.requiredtool('veritysetup') to test_fit_verity_roothash_signed(), matching test_fit_verity.py. Verified locally both ways: with veritysetup hidden from PATH the test now skips instead of failing, and with it present it still passes.
* collect Reviewed-by from Simon Glass
v3: address comments by Simon Glass
* assert fdt_find_regions()'s returned count stays within the region array bound, matching the check fit_config_check_sig() enforces, instead of only checking it is non-zero
* tamper the roothash through buf[digest_off] instead of casting away the const from fdt_getprop()'s return value
* note in a comment that the digest tamper stands in for the whole dm-verity node, since salt lives in the same node and is covered by the same region
* fix test_fit_verity_roothash_signed(), which never actually ran:
its ITS baked the tmpdir prefix into the /incbin/() paths while dtc also resolves incbin paths relative to the .its file's own directory (same as tmpdir here), so the path was searched doubled ("tmpdir/tmpdir/kernel.bin") and mkimage always failed with "No such file or directory". Use bare filenames in the ITS and let dtc resolve them; drop the now-unused %(tmpdir)s substitution.
v2: address comments by Tom Rini
* drop the VISIBLE_IF_UT visibility macro and simply make fit_config_get_signed_nodes() non-static
* convert the comment block in test_fit_verity_sign.py into a module docstring, add an ITS example and add a page under doc/develop/pytest/ so the module is rendered in the generated documentation
--- boot/image-fit-sig.c | 14 +- doc/develop/pytest/test_fit_verity_sign.rst | 10 + include/image.h | 23 +++ test/boot/fit_verity.c | 200 +++++++++++++++++++ test/py/tests/test_fit_verity_sign.py | 203 ++++++++++++++++++++ 5 files changed, 443 insertions(+), 7 deletions(-) create mode 100644 doc/develop/pytest/test_fit_verity_sign.rst create mode 100644 test/py/tests/test_fit_verity_sign.py
boot/image-fit-sig.c +7 -7
@@ -341,7 +341,7 @@ static int fit_config_add_hash(const void *fit, int image_noffset, }/** -* fit_config_get_hash_list() - Build the list of nodes to hash +* fit_config_get_signed_nodes() - Build the list of nodes to hash * * Works through every image referenced by the configuration and collects the * node paths: root + config + all referenced images with their hash,
@@ -358,9 +358,9 @@ static int fit_config_add_hash(const void *fit, int image_noffset,* @buf_len: Size of @buf * Return: number of entries in @node_inc, or -ve on error */ -static int fit_config_get_hash_list (const void *fit,int conf_noffset, -char **node_inc,int max_nodes, -char *buf,int buf_len) +int fit_config_get_signed_nodes (const void *fit,int conf_noffset, +char **node_inc,int max_nodes, +char *buf,int buf_len) {const char *conf_name;int image_count;
@@ -500,9 +500,9 @@ static int fit_config_check_sig(const void *fit, int noffset, int conf_noffset, }/* Build the node list from the config, ignoring hashed-nodes */ - count =fit_config_get_hash_list (fit, conf_noffset, - node_inc,IMAGE_MAX_HASHED_NODES , - hash_buf,sizeof (hash_buf)); + count =fit_config_get_signed_nodes (fit, conf_noffset, + node_inc,IMAGE_MAX_HASHED_NODES , + hash_buf,sizeof (hash_buf));if (count <0 ) { *err_msgp ="Failed to build hash node list" ;return -1 ;
doc/develop/pytest/test_fit_verity_sign.rst new file +10 -0
@@ -0,0 +1,10 @@ +.. SPDX-License-Identifier: GPL-2.0+ + +test_fit_verity_sign +==================== + +.. automodule:: test_fit_verity_sign + :synopsis: + :member-order: bysource + :members: + :undoc-members:
include/image.h +23 -0
@@ -1897,6 +1897,29 @@ struct image_region *fit_region_make_list(const void *fit, struct fdt_region *fdt_regions, int count, struct image_region *region); +/** +* fit_config_get_signed_nodes() - Build the list of nodes covered by a config +* signature +* +* Collects the paths of the nodes that the configuration signature is +* computed over: the root node, the configuration node, and for each image +* referenced by the configuration its node, its hash subnodes and its cipher +* and dm-verity subnodes. The result is the same node list used when creating +* and verifying the signature, and is suitable for passing to +* fdt_find_regions(). +* +* @fit: FIT blob +* @conf_noffset: Configuration node offset +* @node_inc: Array to fill with pointers to packed path strings +* @max_nodes: Number of entries in @node_inc +* @buf: Buffer for the packed null-terminated path strings +* @buf_len: Size of @buf +* Return: number of entries written to @node_inc, or -ve on error +*/ +int fit_config_get_signed_nodes (const void *fit,int conf_noffset, +char **node_inc,int max_nodes, +char *buf,int buf_len); +static inline int fit_image_check_target_arch (const void *fdt,int node) {#ifndef USE_HOSTCC
test/boot/fit_verity.c +200 -0
209 lines not on this page · raw message
test/py/tests/test_fit_verity_sign.py new file +203 -0
203 lines not on this page · raw message
Signature · 1 line
2.55.0
Quoted 8 lines · Daniel Golle
A signed FIT configuration can delegate the integrity of a (potentially large) root filesystem image to the kernel's dm-verity instead of having U-Boot hash the whole payload at boot: the FIT carries a "dm-verity" subnode with the roothash, salt and block parameters, U-Boot passes the roothash to Linux through the dm-mod.create bootargs, and dm-verity then validates the filesystem block by block against it.
[...]
Applied to u-boot/main, thanks!
[1/3] boot: fit: factor out node-path collection in fit_config_add_hash()
commit: ba9ce23d21e3536b7de5e5722f20b8d6be695d3c
[2/3] boot: fit: cover the dm-verity roothash with the config signature
commit: 2601d94691c00e0a05dc61241bb91d680519d49b
[3/3] test: fit: verify dm-verity roothash is covered by the config signature
commit: fe9877c7d9dea740985edd11f7ff583e311568beSignature · 1 line
Tom