OP-TEE February 2026

op-tee@lists.trustedfirmware.org

15 participants
15 discussions

[PATCH v20 0/6] Introduction of a remoteproc tee to load signed firmware

by Arnaud Pouliquen

Main updates from version V19[3]: -------------------------------- The devicetree is now structured as follows: firmware { optee { compatible = "linaro,optee-tz"; method = "smc"; #address-cells = <1>; #size-cells = <0>; rproc-service@0 { compatible = "rproc-service-80a4c275-0a47-4905-8285-1486a9771a08"; reg = <0>; #address-cells = <1>; #size-cells = <0>; status = "okay"; m4: m4@0 { compatible = "st,stm32mp15-m4-tee"; reg = <0>; mboxes = <&ipcc 0>, <&ipcc 1>, <&ipcc 2>; mbox-names = "vq0", "vq1", "shutdown"; memory-region = <&vdev0vring0>, <&m_ipc_shm>, <&mcuram2>, <&vdev0vring1>, <&vdev0buffer>, <&retram>; interrupt-parent = <&exti>; interrupts = <68 1>; status = "okay"; }; }; }; }; As a consequence, this version: - Introduces a new stm32_rproc_tee.c remoteproc driver. Instead of further complicating the existing stm32_rproc.c driver, a dedicated TEE-based driver is added. Both drivers are intended to also support the STM32MP2x Cortex-M33 remote processor in a next step. - Reworks the bindings: - Drop the st,stm32-rproc.yaml updates that were introduced in previous revisions. - Add remoteproc-tee.yaml for the "rproc-service-80a4c275-0a47-4905-8285-1486a9771a08" compatible. - Add st,stm32-rproc-tee.yaml for the "st,stm32mp15-m4-tee" compatible. - Reworks the probing sequence: The m4@0 device is now probed by the remoteproc-tee driver, which itself is instantiated by the TEE (OP-TEE) bus. Main updates from version V18[2]: -------------------------------- - rework documentation for the release_fw ops - rework function documentation in remoteproc_tee.c - replace spinlock by mutex and generalize usage in remoteproc_tee.c Main updates from version V17[1]: -------------------------------- - Fix: warning: EXPORT_SYMBOL() is used, but #include <linux/export.h> is missing More details are available in each patch commit message. [1] https://lore.kernel.org/linux-remoteproc/20250613091650.2337411-1-arnaud.po… [2] https://lore.kernel.org/linux-remoteproc/20250616075530.4106090-1-arnaud.po… [3] https://lore.kernel.org/linux-devicetree/20250625094028.758016-1-arnaud.pou… Tested-on: --------- commit 7d0a66e4bb90 ("Linux 6.18") Description of the feature: -------------------------- This series proposes the implementation of a remoteproc tee driver to communicate with a TEE trusted application responsible for authenticating and loading the remoteproc firmware image in an Arm secure context. 1) Principle: The remoteproc tee driver provides services to communicate with the OP-TEE trusted application running on the Trusted Execution Context (TEE). The trusted application in TEE manages the remote processor lifecycle: - authenticating and loading firmware images, - isolating and securing the remote processor memories, - supporting multi-firmware (e.g., TF-M + Zephyr on a Cortex-M33), - managing the start and stop of the firmware by the TEE. 2) Format of the signed image: Refer to: https://github.com/OP-TEE/optee_os/blob/master/ta/remoteproc/src/remoteproc… 3) OP-TEE trusted application API: Refer to: https://github.com/OP-TEE/optee_os/blob/master/ta/remoteproc/include/ta_rem… 4) OP-TEE signature script Refer to: https://github.com/OP-TEE/optee_os/blob/master/scripts/sign_rproc_fw.py Example of usage: sign_rproc_fw.py --in <fw1.elf> --in <fw2.elf> --out <signed_fw.sign> --key ${OP-TEE_PATH}/keys/default.pem 5) Impact on User space Application No sysfs impact. The user only needs to provide the signed firmware image instead of the ELF image. For more information about the implementation, a presentation is available here (note that the format of the signed image has evolved between the presentation and the integration in OP-TEE). https://resources.linaro.org/en/resource/6c5bGvZwUAjX56fvxthxds Arnaud Pouliquen (6): dt-bindings: firmware: Add TEE remoteproc service binding dt-bindings: remoteproc: Add STM32 TEE-controlled rproc binding remoteproc: core: Introduce rproc_pa_to_va helper remoteproc: Introduce optional release_fw operation remoteproc: Add TEE support remoteproc: stm32: Add TEE-controlled STM32 driver .../arm/firmware/linaro,optee-tz.yaml | 6 + .../bindings/remoteproc/remoteproc-tee.yaml | 47 ++ .../remoteproc/st,stm32-rproc-tee.yaml | 100 +++ drivers/remoteproc/Kconfig | 10 + drivers/remoteproc/Makefile | 3 +- drivers/remoteproc/remoteproc_core.c | 52 ++ drivers/remoteproc/remoteproc_internal.h | 6 + drivers/remoteproc/remoteproc_tee.c | 771 ++++++++++++++++++ drivers/remoteproc/stm32_rproc_tee.c | 526 ++++++++++++ include/linux/remoteproc.h | 6 + include/linux/remoteproc_tee.h | 89 ++ 11 files changed, 1615 insertions(+), 1 deletion(-) create mode 100644 Documentation/devicetree/bindings/remoteproc/remoteproc-tee.yaml create mode 100644 Documentation/devicetree/bindings/remoteproc/st,stm32-rproc-tee.yaml create mode 100644 drivers/remoteproc/remoteproc_tee.c create mode 100644 drivers/remoteproc/stm32_rproc_tee.c create mode 100644 include/linux/remoteproc_tee.h base-commit: 7d0a66e4bb9081d75c82ec4957c50034cb0ea449 -- 2.43.0

2 months, 3 weeks

[PATCH v5] tee: optee: prevent use-after-free when the client exits before the supplicant

by Amirreza Zarrabi

Commit 70b0d6b0a199 ("tee: optee: Fix supplicant wait loop") made the client wait as killable so it can be interrupted during shutdown or after a supplicant crash. This changes the original lifetime expectations: the client task can now terminate while the supplicant is still processing its request. If the client exits first it removes the request from its queue and kfree()s it, while the request ID remains in supp->idr. A subsequent lookup on the supplicant path then dereferences freed memory, leading to a use-after-free. Serialise access to the request with supp->mutex: * Hold supp->mutex in optee_supp_recv() and optee_supp_send() while looking up and touching the request. * Let optee_supp_thrd_req() notice that the client has terminated and signal optee_supp_send() accordingly. With these changes the request cannot be freed while the supplicant still has a reference, eliminating the race. Fixes: 70b0d6b0a199 ("tee: optee: Fix supplicant wait loop") Signed-off-by: Amirreza Zarrabi <amirreza.zarrabi(a)oss.qualcomm.com> --- Changes in v5: - Move idr_alloc() inside the supp->mutex, reported by Yuanjia Yeh. - Link to v4: https://lore.kernel.org/r/20260209-fix-use-after-free-v4-1-7c4c4b02368f@oss… Changes in v4: - Pre-allocate request ID when allocating the request. - Cleanup the loop in optee_supp_recv(). - Update the return value for revoked request from -ENOENT to -EBADF. - Link to v3: https://lore.kernel.org/r/20260128-fix-use-after-free-v3-1-b0786670d927@oss… Changes in v3: - Introduce processed flag instead of -1 for req->id. - Update optee_supp_release() as reported by Michael Wu. - Use mutex instead of guard. - Link to v2: https://lore.kernel.org/r/20250617-fix-use-after-free-v2-1-1fbfafec5917@oss… Changes in v2: - Replace the static variable with a sentinel value. - Fix the issue with returning the popped request to the supplicant. - Link to v1: https://lore.kernel.org/r/20250605-fix-use-after-free-v1-1-a70d23bff248@oss… --- drivers/tee/optee/supp.c | 107 ++++++++++++++++++++++++++++++++--------------- 1 file changed, 74 insertions(+), 33 deletions(-) diff --git a/drivers/tee/optee/supp.c b/drivers/tee/optee/supp.c index d0f397c90242..2386bbd38ce7 100644 --- a/drivers/tee/optee/supp.c +++ b/drivers/tee/optee/supp.c @@ -10,7 +10,11 @@ struct optee_supp_req { struct list_head link; + int id; + bool in_queue; + bool processed; + u32 func; u32 ret; size_t num_params; @@ -19,6 +23,9 @@ struct optee_supp_req { struct completion c; }; +/* It is temporary request used for revoked pending request in supp->idr. */ +#define INVALID_REQ_PTR ((struct optee_supp_req *)ERR_PTR(-EBADF)) + void optee_supp_init(struct optee_supp *supp) { memset(supp, 0, sizeof(*supp)); @@ -39,21 +46,23 @@ void optee_supp_release(struct optee_supp *supp) { int id; struct optee_supp_req *req; - struct optee_supp_req *req_tmp; mutex_lock(&supp->mutex); - /* Abort all request retrieved by supplicant */ + /* Abort all request */ idr_for_each_entry(&supp->idr, req, id) { idr_remove(&supp->idr, id); - req->ret = TEEC_ERROR_COMMUNICATION; - complete(&req->c); - } + /* Skip if request was already marked invalid */ + if (IS_ERR(req)) + continue; - /* Abort all queued requests */ - list_for_each_entry_safe(req, req_tmp, &supp->reqs, link) { - list_del(&req->link); - req->in_queue = false; + /* For queued requests where supplicant has not seen it */ + if (req->in_queue) { + list_del(&req->link); + req->in_queue = false; + } + + req->processed = true; req->ret = TEEC_ERROR_COMMUNICATION; complete(&req->c); } @@ -100,8 +109,16 @@ u32 optee_supp_thrd_req(struct tee_context *ctx, u32 func, size_t num_params, /* Insert the request in the request list */ mutex_lock(&supp->mutex); + req->id = idr_alloc(&supp->idr, req, 1, 0, GFP_KERNEL); + if (req->id < 0) { + mutex_unlock(&supp->mutex); + kfree(req); + return TEEC_ERROR_OUT_OF_MEMORY; + } + list_add_tail(&req->link, &supp->reqs); req->in_queue = true; + req->processed = false; mutex_unlock(&supp->mutex); /* Tell an eventual waiter there's a new request */ @@ -117,21 +134,43 @@ u32 optee_supp_thrd_req(struct tee_context *ctx, u32 func, size_t num_params, if (wait_for_completion_killable(&req->c)) { mutex_lock(&supp->mutex); if (req->in_queue) { + /* Supplicant has not seen this request yet. */ + idr_remove(&supp->idr, req->id); list_del(&req->link); req->in_queue = false; + + ret = TEEC_ERROR_COMMUNICATION; + } else if (req->processed) { + /* + * Supplicant has processed this request. Ignore the + * kill signal for now and submit the result. req is not + * in supp->reqs (removed by supp_pop_entry()) nor in + * supp->idr (removed by supp_pop_req()). + */ + ret = req->ret; + } else { + /* + * Supplicant is in the middle of processing this + * request. Replace req with INVALID_REQ_PTR so that + * the ID remains busy, causing optee_supp_send() to + * fail on the next call to supp_pop_req() with this ID. + */ + idr_replace(&supp->idr, INVALID_REQ_PTR, req->id); + ret = TEEC_ERROR_COMMUNICATION; } + mutex_unlock(&supp->mutex); - req->ret = TEEC_ERROR_COMMUNICATION; + } else { + ret = req->ret; } - ret = req->ret; kfree(req); return ret; } static struct optee_supp_req *supp_pop_entry(struct optee_supp *supp, - int num_params, int *id) + int num_params) { struct optee_supp_req *req; @@ -153,10 +192,6 @@ static struct optee_supp_req *supp_pop_entry(struct optee_supp *supp, return ERR_PTR(-EINVAL); } - *id = idr_alloc(&supp->idr, req, 1, 0, GFP_KERNEL); - if (*id < 0) - return ERR_PTR(-ENOMEM); - list_del(&req->link); req->in_queue = false; @@ -214,7 +249,6 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, struct optee *optee = tee_get_drvdata(teedev); struct optee_supp *supp = &optee->supp; struct optee_supp_req *req = NULL; - int id; size_t num_meta; int rc; @@ -224,15 +258,11 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, while (true) { mutex_lock(&supp->mutex); - req = supp_pop_entry(supp, *num_params - num_meta, &id); + req = supp_pop_entry(supp, *num_params - num_meta); + if (req) + break; /* Keep mutex held. */ mutex_unlock(&supp->mutex); - if (req) { - if (IS_ERR(req)) - return PTR_ERR(req); - break; - } - /* * If we didn't get a request we'll block in * wait_for_completion() to avoid needless spinning. @@ -245,6 +275,13 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, return -ERESTARTSYS; } + /* supp->mutex held and req != NULL. */ + + if (IS_ERR(req)) { + mutex_unlock(&supp->mutex); + return PTR_ERR(req); + } + if (num_meta) { /* * tee-supplicant support meta parameters -> requsts can be @@ -252,13 +289,11 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, */ param->attr = TEE_IOCTL_PARAM_ATTR_TYPE_VALUE_INOUT | TEE_IOCTL_PARAM_ATTR_META; - param->u.value.a = id; + param->u.value.a = req->id; param->u.value.b = 0; param->u.value.c = 0; } else { - mutex_lock(&supp->mutex); - supp->req_id = id; - mutex_unlock(&supp->mutex); + supp->req_id = req->id; } *func = req->func; @@ -266,6 +301,7 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, memcpy(param + num_meta, req->param, sizeof(struct tee_param) * req->num_params); + mutex_unlock(&supp->mutex); return 0; } @@ -297,12 +333,17 @@ static struct optee_supp_req *supp_pop_req(struct optee_supp *supp, if (!req) return ERR_PTR(-ENOENT); + /* optee_supp_thrd_req() already returned to optee. */ + if (IS_ERR(req)) + goto failed_req; + if ((num_params - nm) != req->num_params) return ERR_PTR(-EINVAL); + *num_meta = nm; +failed_req: idr_remove(&supp->idr, id); supp->req_id = -1; - *num_meta = nm; return req; } @@ -328,10 +369,9 @@ int optee_supp_send(struct tee_context *ctx, u32 ret, u32 num_params, mutex_lock(&supp->mutex); req = supp_pop_req(supp, num_params, param, &num_meta); - mutex_unlock(&supp->mutex); - if (IS_ERR(req)) { - /* Something is wrong, let supplicant restart. */ + mutex_unlock(&supp->mutex); + /* Something is wrong, let supplicant handel it. */ return PTR_ERR(req); } @@ -355,9 +395,10 @@ int optee_supp_send(struct tee_context *ctx, u32 ret, u32 num_params, } } req->ret = ret; - + req->processed = true; /* Let the requesting thread continue */ complete(&req->c); + mutex_unlock(&supp->mutex); return 0; } --- base-commit: 3f24e4edcd1b8981c6b448ea2680726dedd87279 change-id: 20250604-fix-use-after-free-8ff1b5d5d774 Best regards, -- Amirreza Zarrabi <amirreza.zarrabi(a)oss.qualcomm.com>

2 months, 3 weeks

[PATCH v2] tee: shm: fix slab page refcounting

by Marco Felsch

Skip manipulating the refcount in case of slab pages according commit b9c0e49abfca ("mm: decline to manipulate the refcount on a slab page"). Fixes: b9c0e49abfca ("mm: decline to manipulate the refcount on a slab page") Signed-off-by: Marco Felsch <m.felsch(a)pengutronix.de> --- v2: - Make use of page variable v1: - https://lore.kernel.org/all/20250325195021.3589797-1-m.felsch@pengutronix.d… drivers/tee/tee_shm.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) diff --git a/drivers/tee/tee_shm.c b/drivers/tee/tee_shm.c index daf6e5cfd59a..35f0ac359b12 100644 --- a/drivers/tee/tee_shm.c +++ b/drivers/tee/tee_shm.c @@ -19,16 +19,24 @@ static void shm_put_kernel_pages(struct page **pages, size_t page_count) { size_t n; - for (n = 0; n < page_count; n++) - put_page(pages[n]); + for (n = 0; n < page_count; n++) { + struct page *page = pages[n]; + + if (!PageSlab(page)) + put_page(page); + } } static void shm_get_kernel_pages(struct page **pages, size_t page_count) { size_t n; - for (n = 0; n < page_count; n++) - get_page(pages[n]); + for (n = 0; n < page_count; n++) { + struct page *page = pages[n]; + + if (!PageSlab(page)) + get_page(page); + } } static void release_registered_pages(struct tee_shm *shm) -- 2.39.5

2 months, 3 weeks

[PATCH v4] tee: optee: prevent use-after-free when the client exits before the supplicant

by Amirreza Zarrabi

Commit 70b0d6b0a199 ("tee: optee: Fix supplicant wait loop") made the client wait as killable so it can be interrupted during shutdown or after a supplicant crash. This changes the original lifetime expectations: the client task can now terminate while the supplicant is still processing its request. If the client exits first it removes the request from its queue and kfree()s it, while the request ID remains in supp->idr. A subsequent lookup on the supplicant path then dereferences freed memory, leading to a use-after-free. Serialise access to the request with supp->mutex: * Hold supp->mutex in optee_supp_recv() and optee_supp_send() while looking up and touching the request. * Let optee_supp_thrd_req() notice that the client has terminated and signal optee_supp_send() accordingly. With these changes the request cannot be freed while the supplicant still has a reference, eliminating the race. Fixes: 70b0d6b0a199 ("tee: optee: Fix supplicant wait loop") Signed-off-by: Amirreza Zarrabi <amirreza.zarrabi(a)oss.qualcomm.com> --- Changes in v4: - Pre-allocate request ID when allocating the request. - Cleanup the loop in optee_supp_recv(). - Update the return value for revoked request from -ENOENT to -EBADF. - Link to v3: https://lore.kernel.org/r/20260128-fix-use-after-free-v3-1-b0786670d927@oss… Changes in v3: - Introduce processed flag instead of -1 for req->id. - Update optee_supp_release() as reported by Michael Wu. - Use mutex instead of guard. - Link to v2: https://lore.kernel.org/r/20250617-fix-use-after-free-v2-1-1fbfafec5917@oss… Changes in v2: - Replace the static variable with a sentinel value. - Fix the issue with returning the popped request to the supplicant. - Link to v1: https://lore.kernel.org/r/20250605-fix-use-after-free-v1-1-a70d23bff248@oss… --- drivers/tee/optee/supp.c | 106 ++++++++++++++++++++++++++++++++--------------- 1 file changed, 73 insertions(+), 33 deletions(-) diff --git a/drivers/tee/optee/supp.c b/drivers/tee/optee/supp.c index d0f397c90242..c1ae76df7067 100644 --- a/drivers/tee/optee/supp.c +++ b/drivers/tee/optee/supp.c @@ -10,7 +10,11 @@ struct optee_supp_req { struct list_head link; + int id; + bool in_queue; + bool processed; + u32 func; u32 ret; size_t num_params; @@ -19,6 +23,9 @@ struct optee_supp_req { struct completion c; }; +/* It is temporary request used for revoked pending request in supp->idr. */ +#define INVALID_REQ_PTR ((struct optee_supp_req *)ERR_PTR(-EBADF)) + void optee_supp_init(struct optee_supp *supp) { memset(supp, 0, sizeof(*supp)); @@ -39,21 +46,23 @@ void optee_supp_release(struct optee_supp *supp) { int id; struct optee_supp_req *req; - struct optee_supp_req *req_tmp; mutex_lock(&supp->mutex); - /* Abort all request retrieved by supplicant */ + /* Abort all request */ idr_for_each_entry(&supp->idr, req, id) { idr_remove(&supp->idr, id); - req->ret = TEEC_ERROR_COMMUNICATION; - complete(&req->c); - } + /* Skip if request was already marked invalid */ + if (IS_ERR(req)) + continue; - /* Abort all queued requests */ - list_for_each_entry_safe(req, req_tmp, &supp->reqs, link) { - list_del(&req->link); - req->in_queue = false; + /* For queued requests where supplicant has not seen it */ + if (req->in_queue) { + list_del(&req->link); + req->in_queue = false; + } + + req->processed = true; req->ret = TEEC_ERROR_COMMUNICATION; complete(&req->c); } @@ -93,6 +102,12 @@ u32 optee_supp_thrd_req(struct tee_context *ctx, u32 func, size_t num_params, if (!req) return TEEC_ERROR_OUT_OF_MEMORY; + req->id = idr_alloc(&supp->idr, req, 1, 0, GFP_KERNEL); + if (req->id < 0) { + kfree(req); + return TEEC_ERROR_OUT_OF_MEMORY; + } + init_completion(&req->c); req->func = func; req->num_params = num_params; @@ -102,6 +117,7 @@ u32 optee_supp_thrd_req(struct tee_context *ctx, u32 func, size_t num_params, mutex_lock(&supp->mutex); list_add_tail(&req->link, &supp->reqs); req->in_queue = true; + req->processed = false; mutex_unlock(&supp->mutex); /* Tell an eventual waiter there's a new request */ @@ -117,21 +133,43 @@ u32 optee_supp_thrd_req(struct tee_context *ctx, u32 func, size_t num_params, if (wait_for_completion_killable(&req->c)) { mutex_lock(&supp->mutex); if (req->in_queue) { + /* Supplicant has not seen this request yet. */ + idr_remove(&supp->idr, req->id); list_del(&req->link); req->in_queue = false; + + ret = TEEC_ERROR_COMMUNICATION; + } else if (req->processed) { + /* + * Supplicant has processed this request. Ignore the + * kill signal for now and submit the result. req is not + * in supp->reqs (removed by supp_pop_entry()) nor in + * supp->idr (removed by supp_pop_req()). + */ + ret = req->ret; + } else { + /* + * Supplicant is in the middle of processing this + * request. Replace req with INVALID_REQ_PTR so that + * the ID remains busy, causing optee_supp_send() to + * fail on the next call to supp_pop_req() with this ID. + */ + idr_replace(&supp->idr, INVALID_REQ_PTR, req->id); + ret = TEEC_ERROR_COMMUNICATION; } + mutex_unlock(&supp->mutex); - req->ret = TEEC_ERROR_COMMUNICATION; + } else { + ret = req->ret; } - ret = req->ret; kfree(req); return ret; } static struct optee_supp_req *supp_pop_entry(struct optee_supp *supp, - int num_params, int *id) + int num_params) { struct optee_supp_req *req; @@ -153,10 +191,6 @@ static struct optee_supp_req *supp_pop_entry(struct optee_supp *supp, return ERR_PTR(-EINVAL); } - *id = idr_alloc(&supp->idr, req, 1, 0, GFP_KERNEL); - if (*id < 0) - return ERR_PTR(-ENOMEM); - list_del(&req->link); req->in_queue = false; @@ -214,7 +248,6 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, struct optee *optee = tee_get_drvdata(teedev); struct optee_supp *supp = &optee->supp; struct optee_supp_req *req = NULL; - int id; size_t num_meta; int rc; @@ -224,15 +257,11 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, while (true) { mutex_lock(&supp->mutex); - req = supp_pop_entry(supp, *num_params - num_meta, &id); + req = supp_pop_entry(supp, *num_params - num_meta); + if (req) + break; /* Keep mutex held. */ mutex_unlock(&supp->mutex); - if (req) { - if (IS_ERR(req)) - return PTR_ERR(req); - break; - } - /* * If we didn't get a request we'll block in * wait_for_completion() to avoid needless spinning. @@ -245,6 +274,13 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, return -ERESTARTSYS; } + /* supp->mutex held and req != NULL. */ + + if (IS_ERR(req)) { + mutex_unlock(&supp->mutex); + return PTR_ERR(req); + } + if (num_meta) { /* * tee-supplicant support meta parameters -> requsts can be @@ -252,13 +288,11 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, */ param->attr = TEE_IOCTL_PARAM_ATTR_TYPE_VALUE_INOUT | TEE_IOCTL_PARAM_ATTR_META; - param->u.value.a = id; + param->u.value.a = req->id; param->u.value.b = 0; param->u.value.c = 0; } else { - mutex_lock(&supp->mutex); - supp->req_id = id; - mutex_unlock(&supp->mutex); + supp->req_id = req->id; } *func = req->func; @@ -266,6 +300,7 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, memcpy(param + num_meta, req->param, sizeof(struct tee_param) * req->num_params); + mutex_unlock(&supp->mutex); return 0; } @@ -297,12 +332,17 @@ static struct optee_supp_req *supp_pop_req(struct optee_supp *supp, if (!req) return ERR_PTR(-ENOENT); + /* optee_supp_thrd_req() already returned to optee. */ + if (IS_ERR(req)) + goto failed_req; + if ((num_params - nm) != req->num_params) return ERR_PTR(-EINVAL); + *num_meta = nm; +failed_req: idr_remove(&supp->idr, id); supp->req_id = -1; - *num_meta = nm; return req; } @@ -328,10 +368,9 @@ int optee_supp_send(struct tee_context *ctx, u32 ret, u32 num_params, mutex_lock(&supp->mutex); req = supp_pop_req(supp, num_params, param, &num_meta); - mutex_unlock(&supp->mutex); - if (IS_ERR(req)) { - /* Something is wrong, let supplicant restart. */ + mutex_unlock(&supp->mutex); + /* Something is wrong, let supplicant handel it. */ return PTR_ERR(req); } @@ -355,9 +394,10 @@ int optee_supp_send(struct tee_context *ctx, u32 ret, u32 num_params, } } req->ret = ret; - + req->processed = true; /* Let the requesting thread continue */ complete(&req->c); + mutex_unlock(&supp->mutex); return 0; } --- base-commit: 3f24e4edcd1b8981c6b448ea2680726dedd87279 change-id: 20250604-fix-use-after-free-8ff1b5d5d774 Best regards, -- Amirreza Zarrabi <amirreza.zarrabi(a)oss.qualcomm.com>

3 months

[PATCH v3] tee: optee: prevent use-after-free when the client exits before the supplicant

by Amirreza Zarrabi

Commit 70b0d6b0a199 ("tee: optee: Fix supplicant wait loop") made the client wait as killable so it can be interrupted during shutdown or after a supplicant crash. This changes the original lifetime expectations: the client task can now terminate while the supplicant is still processing its request. If the client exits first it removes the request from its queue and kfree()s it, while the request ID remains in supp->idr. A subsequent lookup on the supplicant path then dereferences freed memory, leading to a use-after-free. Serialise access to the request with supp->mutex: * Hold supp->mutex in optee_supp_recv() and optee_supp_send() while looking up and touching the request. * Let optee_supp_thrd_req() notice that the client has terminated and signal optee_supp_send() accordingly. With these changes the request cannot be freed while the supplicant still has a reference, eliminating the race. Fixes: 70b0d6b0a199 ("tee: optee: Fix supplicant wait loop") Signed-off-by: Amirreza Zarrabi <amirreza.zarrabi(a)oss.qualcomm.com> --- Changes in v3: - Introduce processed flag instead of -1 for req->id. - Update optee_supp_release() as reported by Michael Wu. - Use mutex instead of guard. - Link to v2: https://lore.kernel.org/r/20250617-fix-use-after-free-v2-1-1fbfafec5917@oss… Changes in v2: - Replace the static variable with a sentinel value. - Fix the issue with returning the popped request to the supplicant. - Link to v1: https://lore.kernel.org/r/20250605-fix-use-after-free-v1-1-a70d23bff248@oss… --- drivers/tee/optee/supp.c | 122 +++++++++++++++++++++++++++++++++-------------- 1 file changed, 86 insertions(+), 36 deletions(-) diff --git a/drivers/tee/optee/supp.c b/drivers/tee/optee/supp.c index d0f397c90242..0ec66008df19 100644 --- a/drivers/tee/optee/supp.c +++ b/drivers/tee/optee/supp.c @@ -10,7 +10,11 @@ struct optee_supp_req { struct list_head link; + int id; + bool in_queue; + bool processed; + u32 func; u32 ret; size_t num_params; @@ -19,6 +23,9 @@ struct optee_supp_req { struct completion c; }; +/* It is temporary request used for invalid pending request in supp->idr. */ +#define INVALID_REQ_PTR ((struct optee_supp_req *)ERR_PTR(-ENOENT)) + void optee_supp_init(struct optee_supp *supp) { memset(supp, 0, sizeof(*supp)); @@ -46,6 +53,10 @@ void optee_supp_release(struct optee_supp *supp) /* Abort all request retrieved by supplicant */ idr_for_each_entry(&supp->idr, req, id) { idr_remove(&supp->idr, id); + /* Skip if request was already marked invalid */ + if (IS_ERR(req)) + continue; + req->ret = TEEC_ERROR_COMMUNICATION; complete(&req->c); } @@ -102,6 +113,7 @@ u32 optee_supp_thrd_req(struct tee_context *ctx, u32 func, size_t num_params, mutex_lock(&supp->mutex); list_add_tail(&req->link, &supp->reqs); req->in_queue = true; + req->processed = false; mutex_unlock(&supp->mutex); /* Tell an eventual waiter there's a new request */ @@ -117,21 +129,40 @@ u32 optee_supp_thrd_req(struct tee_context *ctx, u32 func, size_t num_params, if (wait_for_completion_killable(&req->c)) { mutex_lock(&supp->mutex); if (req->in_queue) { + /* Supplicant has not seen this request yet. */ list_del(&req->link); req->in_queue = false; + + ret = TEEC_ERROR_COMMUNICATION; + } else if (req->processed) { + /* + * Supplicant has processed this request. Ignore the + * kill signal for now and submit the result. + */ + ret = req->ret; + } else { + /* + * Supplicant is in the middle of processing this + * request. Replace req with INVALID_REQ_PTR so that + * the ID remains busy, causing optee_supp_send() to + * fail on the next call to supp_pop_req() with this ID. + */ + idr_replace(&supp->idr, INVALID_REQ_PTR, req->id); + ret = TEEC_ERROR_COMMUNICATION; } + mutex_unlock(&supp->mutex); - req->ret = TEEC_ERROR_COMMUNICATION; + } else { + ret = req->ret; } - ret = req->ret; kfree(req); return ret; } static struct optee_supp_req *supp_pop_entry(struct optee_supp *supp, - int num_params, int *id) + int num_params) { struct optee_supp_req *req; @@ -153,8 +184,8 @@ static struct optee_supp_req *supp_pop_entry(struct optee_supp *supp, return ERR_PTR(-EINVAL); } - *id = idr_alloc(&supp->idr, req, 1, 0, GFP_KERNEL); - if (*id < 0) + req->id = idr_alloc(&supp->idr, req, 1, 0, GFP_KERNEL); + if (req->id < 0) return ERR_PTR(-ENOMEM); list_del(&req->link); @@ -214,7 +245,6 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, struct optee *optee = tee_get_drvdata(teedev); struct optee_supp *supp = &optee->supp; struct optee_supp_req *req = NULL; - int id; size_t num_meta; int rc; @@ -224,15 +254,48 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, while (true) { mutex_lock(&supp->mutex); - req = supp_pop_entry(supp, *num_params - num_meta, &id); - mutex_unlock(&supp->mutex); - if (req) { - if (IS_ERR(req)) - return PTR_ERR(req); - break; + req = supp_pop_entry(supp, *num_params - num_meta); + if (!req) { + mutex_unlock(&supp->mutex); + goto wait_for_request; + } + + if (IS_ERR(req)) { + rc = PTR_ERR(req); + mutex_unlock(&supp->mutex); + + return rc; } + /* + * Process the request while holding the lock, so that + * optee_supp_thrd_req() doesn't pull the request from under us. + */ + + if (num_meta) { + /* + * tee-supplicant support meta parameters -> + * requests can be processed asynchronously. + */ + param->attr = TEE_IOCTL_PARAM_ATTR_TYPE_VALUE_INOUT | + TEE_IOCTL_PARAM_ATTR_META; + param->u.value.a = req->id; + param->u.value.b = 0; + param->u.value.c = 0; + } else { + supp->req_id = req->id; + } + + *func = req->func; + *num_params = req->num_params + num_meta; + memcpy(param + num_meta, req->param, + sizeof(struct tee_param) * req->num_params); + + mutex_unlock(&supp->mutex); + return 0; + +wait_for_request: /* * If we didn't get a request we'll block in * wait_for_completion() to avoid needless spinning. @@ -243,29 +306,10 @@ int optee_supp_recv(struct tee_context *ctx, u32 *func, u32 *num_params, */ if (wait_for_completion_interruptible(&supp->reqs_c)) return -ERESTARTSYS; - } - if (num_meta) { - /* - * tee-supplicant support meta parameters -> requsts can be - * processed asynchronously. - */ - param->attr = TEE_IOCTL_PARAM_ATTR_TYPE_VALUE_INOUT | - TEE_IOCTL_PARAM_ATTR_META; - param->u.value.a = id; - param->u.value.b = 0; - param->u.value.c = 0; - } else { - mutex_lock(&supp->mutex); - supp->req_id = id; - mutex_unlock(&supp->mutex); + /* Check for the next request in the queue. */ } - *func = req->func; - *num_params = req->num_params + num_meta; - memcpy(param + num_meta, req->param, - sizeof(struct tee_param) * req->num_params); - return 0; } @@ -297,12 +341,18 @@ static struct optee_supp_req *supp_pop_req(struct optee_supp *supp, if (!req) return ERR_PTR(-ENOENT); + /* optee_supp_thrd_req() already returned to optee. */ + if (IS_ERR(req)) + goto failed_req; + if ((num_params - nm) != req->num_params) return ERR_PTR(-EINVAL); + *num_meta = nm; +failed_req: idr_remove(&supp->idr, id); supp->req_id = -1; - *num_meta = nm; + return req; } @@ -328,9 +378,8 @@ int optee_supp_send(struct tee_context *ctx, u32 ret, u32 num_params, mutex_lock(&supp->mutex); req = supp_pop_req(supp, num_params, param, &num_meta); - mutex_unlock(&supp->mutex); - if (IS_ERR(req)) { + mutex_unlock(&supp->mutex); /* Something is wrong, let supplicant restart. */ return PTR_ERR(req); } @@ -355,9 +404,10 @@ int optee_supp_send(struct tee_context *ctx, u32 ret, u32 num_params, } } req->ret = ret; - + req->processed = true; /* Let the requesting thread continue */ complete(&req->c); + mutex_unlock(&supp->mutex); return 0; } --- base-commit: 3f24e4edcd1b8981c6b448ea2680726dedd87279 change-id: 20250604-fix-use-after-free-8ff1b5d5d774 Best regards, -- Amirreza Zarrabi <amirreza.zarrabi(a)oss.qualcomm.com>

3 months

2026

2025

2024

2023

2022

2021

2020

OP-TEE February 2026