{ "dataType": "CVE_RECORD", "dataVersion": "5.2", "cveMetadata": { "cveId": "CVE-2024-26623", "assignerOrgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67", "state": "PUBLISHED", "assignerShortName": "Linux", "dateReserved": "2024-02-19T14:20:24.134Z", "datePublished": "2024-03-06T06:45:32.278Z", "dateUpdated": "2026-08-05T11:26:20.441Z" }, "containers": { "cna": { "providerMetadata": { "orgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67", "shortName": "Linux", "dateUpdated": "2026-08-05T11:26:20.441Z" }, "descriptions": [ { "lang": "en", "value": "In the Linux kernel, the following vulnerability has been resolved:\n\npds_core: Prevent race issues involving the adminq\n\nThere are multiple paths that can result in using the pdsc's\nadminq.\n\n[1] pdsc_adminq_isr and the resulting work from queue_work(),\n i.e. pdsc_work_thread()->pdsc_process_adminq()\n\n[2] pdsc_adminq_post()\n\nWhen the device goes through reset via PCIe reset and/or\na fw_down/fw_up cycle due to bad PCIe state or bad device\nstate the adminq is destroyed and recreated.\n\nA NULL pointer dereference can happen if [1] or [2] happens\nafter the adminq is already destroyed.\n\nIn order to fix this, add some further state checks and\nimplement reference counting for adminq uses. Reference\ncounting was used because multiple threads can attempt to\naccess the adminq at the same time via [1] or [2]. Additionally,\nmultiple clients (i.e. pds-vfio-pci) can be using [2]\nat the same time.\n\nThe adminq_refcnt is initialized to 1 when the adminq has been\nallocated and is ready to use. Users/clients of the adminq\n(i.e. [1] and [2]) will increment the refcnt when they are using\nthe adminq. When the driver goes into a fw_down cycle it will\nset the PDSC_S_FW_DEAD bit and then wait for the adminq_refcnt\nto hit 1. Setting the PDSC_S_FW_DEAD before waiting will prevent\nany further adminq_refcnt increments. Waiting for the\nadminq_refcnt to hit 1 allows for any current users of the adminq\nto finish before the driver frees the adminq. Once the\nadminq_refcnt hits 1 the driver clears the refcnt to signify that\nthe adminq is deleted and cannot be used. On the fw_up cycle the\ndriver will once again initialize the adminq_refcnt to 1 allowing\nthe adminq to be used again." } ], "metrics": [ { "cvssV3_1": { "version": "3.1", "vectorString": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H", "baseScore": 7.8, "baseSeverity": "HIGH" }, "scenarios": [ { "lang": "en", "value": "AV:L - The adminq is driven from local interfaces only — vhost-vdpa ioctls via drivers/vdpa/pds, VFIO migration commands via drivers/vfio/pci/pds/cmds.c, and devlink/PCI reset paths; no remote peer data reaches this code.\nAC:L - A local attacker controls one side of the race directly (unbounded concurrent pdsc_adminq_post() calls from multiple client threads) and can influence the teardown side, since command timeouts/ENXIO queue health_work and a VFIO/vDPA holder can induce device resets, so the window is repeatedly retryable at will.\nPR:L - An unprivileged local user holding a vhost-vdpa or VFIO device fd — a normal delegation to a VMM or container in cloud/virtualization deployments where these AMD Pensando cards are used — can post adminq commands; no root or init-namespace capability is required.\nUI:N - Exploitation only requires the attacker's own concurrent adminq activity across a firmware-recovery or PCIe reset cycle; no victim action is involved.\nS:U - The corruption and crash occur in the host kernel that owns the driver; no VM, IOMMU, or sandbox boundary is crossed.\nC:H - Racing pdsc_process_notifyq()/pdsc_process_adminq() read cq->info and the completion buffer after vfree()/dma_free_coherent(), so freed and potentially reallocated kernel memory is read and its contents propagated to client drivers through pdsc_notify(), giving a use-after-free read primitive.\nI:H - __pdsc_adminq_post() writes q_info->wc, q_info->dest and a 64-byte partly attacker-controlled memcpy into vfree'd descriptor memory and freed DMA-coherent pages, and pdsc_process_adminq() calls complete_all() through a pointer taken from freed memory — a use-after-free write usable for heap grooming and control-flow corruption.\nA:H - The commit states a NULL pointer dereference occurs when the adminq is used after destruction; additionally intr_ctrl/info_regs are dereferenced after iounmap and kern_dbpage is NULL, each producing a kernel oops or panic." } ] } ], "affected": [ { "product": "Linux", "vendor": "Linux", "defaultStatus": "unaffected", "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git", "programFiles": [ "drivers/net/ethernet/amd/pds_core/adminq.c", "drivers/net/ethernet/amd/pds_core/core.c", "drivers/net/ethernet/amd/pds_core/core.h" ], "versions": [ { "version": "01ba61b55b2041a39c54aefb3153c770dd59a0ef", "lessThan": "22cd6046eb2148b18990257505834dd45c672a1b", "status": "affected", "versionType": "git" }, { "version": "01ba61b55b2041a39c54aefb3153c770dd59a0ef", "lessThan": "5939feb63ea1f011027576c64b68b681cbad31ca", "status": "affected", "versionType": "git" }, { "version": "01ba61b55b2041a39c54aefb3153c770dd59a0ef", "lessThan": "7e82a8745b951b1e794cc780d46f3fbee5e93447", "status": "affected", "versionType": "git" } ] }, { "product": "Linux", "vendor": "Linux", "defaultStatus": "affected", "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git", "programFiles": [ "drivers/net/ethernet/amd/pds_core/adminq.c", "drivers/net/ethernet/amd/pds_core/core.c", "drivers/net/ethernet/amd/pds_core/core.h" ], "versions": [ { "version": "6.4", "status": "affected" }, { "version": "0", "lessThan": "6.4", "status": "unaffected", "versionType": "semver" }, { "version": "6.6.16", "lessThanOrEqual": "6.6.*", "status": "unaffected", "versionType": "semver" }, { "version": "6.7.4", "lessThanOrEqual": "6.7.*", "status": "unaffected", "versionType": "semver" }, { "version": "6.8", "lessThanOrEqual": "*", "status": "unaffected", "versionType": "original_commit_for_fix" } ] } ], "cpeApplicability": [ { "nodes": [ { "operator": "OR", "negate": false, "cpeMatch": [ { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "6.4", "versionEndExcluding": "6.6.16" }, { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "6.4", "versionEndExcluding": "6.7.4" }, { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "6.4", "versionEndExcluding": "6.8" } ] } ] } ], "references": [ { "url": "https://git.kernel.org/stable/c/22cd6046eb2148b18990257505834dd45c672a1b" }, { "url": "https://git.kernel.org/stable/c/5939feb63ea1f011027576c64b68b681cbad31ca" }, { "url": "https://git.kernel.org/stable/c/7e82a8745b951b1e794cc780d46f3fbee5e93447" } ], "title": "pds_core: Prevent race issues involving the adminq", "x_generator": { "engine": "bippy-1.2.0" } }, "adp": [ { "title": "CISA ADP Vulnrichment", "metrics": [ { "other": { "type": "ssvc", "content": { "id": "CVE-2024-26623", "role": "CISA Coordinator", "options": [ { "Exploitation": "none" }, { "Automatable": "no" }, { "Technical Impact": "partial" } ], "version": "2.0.3", "timestamp": "2024-03-06T14:15:26.675181Z" } } } ], "providerMetadata": { "orgId": "134c704f-9b21-4f2e-91b3-4a467353bcc0", "shortName": "CISA-ADP", "dateUpdated": "2024-06-04T17:48:43.755Z" } }, { "providerMetadata": { "orgId": "af854a3a-2127-422b-91ae-364da2661108", "shortName": "CVE", "dateUpdated": "2024-08-02T00:07:19.860Z" }, "title": "CVE Program Container", "references": [ { "url": "https://git.kernel.org/stable/c/22cd6046eb2148b18990257505834dd45c672a1b", "tags": [ "x_transferred" ] }, { "url": "https://git.kernel.org/stable/c/5939feb63ea1f011027576c64b68b681cbad31ca", "tags": [ "x_transferred" ] }, { "url": "https://git.kernel.org/stable/c/7e82a8745b951b1e794cc780d46f3fbee5e93447", "tags": [ "x_transferred" ] } ] } ] } }