2023-03-17 17:29:09

by Etienne Carriere

[permalink] [raw]
Subject: [PATCH v5 1/2] dt-bindings: optee driver interrupt can be a per-cpu interrupt

Explicit in optee firmware device tree bindings that the interrupt
used by optee driver for async notification can be a peripheral
interrupt or a per-cpu interrupt.

Signed-off-by: Etienne Carriere <[email protected]>
---
No change since v4

Changes since v3:
- Patch added in this v4 to address review comments.
---
.../devicetree/bindings/arm/firmware/linaro,optee-tz.yaml | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/Documentation/devicetree/bindings/arm/firmware/linaro,optee-tz.yaml b/Documentation/devicetree/bindings/arm/firmware/linaro,optee-tz.yaml
index d4dc0749f9fd..5d033570b57b 100644
--- a/Documentation/devicetree/bindings/arm/firmware/linaro,optee-tz.yaml
+++ b/Documentation/devicetree/bindings/arm/firmware/linaro,optee-tz.yaml
@@ -28,7 +28,8 @@ properties:
maxItems: 1
description: |
This interrupt which is used to signal an event by the secure world
- software is expected to be edge-triggered.
+ software is expected to be either a per-cpu interrupt or an
+ edge-triggered peripheral interrupt.

method:
enum: [smc, hvc]
--
2.25.1



2023-03-17 17:29:17

by Etienne Carriere

[permalink] [raw]
Subject: [PATCH v5 2/2] optee: add per cpu asynchronous notification

Implements use of per-cpu irq for optee asynchronous notification.

Existing optee async notif implementation allows OP-TEE world to
raise an interrupt on which Linux optee driver will query some pending
events. This change allows the signaling interrupt to be a per-cpu
interrupt as with Arm GIC PPIs. Using a PPI instead of an SPI is useful
when no GIC lines are provisioned in the chip design and there are spare
PPI lines.

Cc: Jens Wiklander <[email protected]>
Cc: Sumit Garg <[email protected]>
Cc: Marc Zyngier <[email protected]>

Co-developed-by: Alexandre Torgue <[email protected]>
Signed-off-by: Alexandre Torgue <[email protected]>
Signed-off-by: Etienne Carriere <[email protected]>
---
Changes since v4:
- Used cpuhp_setup_state()/cpuhp_remove_state() to enable/disable the
per-cpu interrupt and let CPU hotplug also manage it.
- Added a call to destroy_workqueue() that was missing in uninit_pcpu_irq().
- Rephrased commit message.

Changes since v3:
- Fixed typo in commit message.
- Added few words in commit message about why we do this change.
- Appended a 2nd commit to the series for request DT bindings update.

Changes since v2:
- Irq and per-cpu irq no more share the primary same handler function
but have a common irq_handler() helper function.
- Removed useless spinlocks.
- Wrapped lines > 80 char.

Changes since v1:
- Fixed missing __percpu attribute reported by kernel test robot.
- Rephrased commit message and added Cc tags.
---
drivers/tee/optee/optee_private.h | 24 ++++++
drivers/tee/optee/smc_abi.c | 125 +++++++++++++++++++++++++++++-
2 files changed, 145 insertions(+), 4 deletions(-)

diff --git a/drivers/tee/optee/optee_private.h b/drivers/tee/optee/optee_private.h
index 04ae58892608..6129d43dd354 100644
--- a/drivers/tee/optee/optee_private.h
+++ b/drivers/tee/optee/optee_private.h
@@ -94,11 +94,35 @@ struct optee_supp {
struct completion reqs_c;
};

+/*
+ * struct optee_pcpu - per cpu notif private struct passed to work functions
+ * @optee optee device reference
+ */
+struct optee_pcpu {
+ struct optee *optee;
+};
+
+/*
+ * struct optee_smc - optee smc communication struct
+ * @invoke_fn handler function to invoke secure monitor
+ * @memremaped_shm virtual address of memory in shared memory pool
+ * @sec_caps: secure world capabilities defined by
+ * OPTEE_SMC_SEC_CAP_* in optee_smc.h
+ * @notif_irq interrupt used as async notification by OP-TEE or 0
+ * @optee_pcpu per_cpu optee instance for per cpu work or NULL
+ * @notif_pcpu_wq workqueue for per cpu aynchronous notification or NULL
+ * @notif_pcpu_work work for per cpu asynchronous notification
+ * @notif_cpuhp_state CPU hotplug state assigned for pcpu interrupt management
+ */
struct optee_smc {
optee_invoke_fn *invoke_fn;
void *memremaped_shm;
u32 sec_caps;
unsigned int notif_irq;
+ struct optee_pcpu __percpu *optee_pcpu;
+ struct workqueue_struct *notif_pcpu_wq;
+ struct work_struct notif_pcpu_work;
+ unsigned int notif_cpuhp_state;
};

/**
diff --git a/drivers/tee/optee/smc_abi.c b/drivers/tee/optee/smc_abi.c
index a1c1fa1a9c28..a2473fbe2461 100644
--- a/drivers/tee/optee/smc_abi.c
+++ b/drivers/tee/optee/smc_abi.c
@@ -52,6 +52,23 @@
*/
#define OPTEE_MIN_STATIC_POOL_ALIGN 9 /* 512 bytes aligned */

+/* SMC ABI considers at most a single TEE firmware */
+static unsigned int pcpu_irq_num;
+
+static int optee_cpuhp_enable_pcpu_irq(unsigned int cpu)
+{
+ enable_percpu_irq(pcpu_irq_num, IRQ_TYPE_NONE);
+
+ return 0;
+}
+
+static int optee_cpuhp_disable_pcpu_irq(unsigned int cpu)
+{
+ disable_percpu_irq(pcpu_irq_num);
+
+ return 0;
+}
+
/*
* 1. Convert between struct tee_param and struct optee_msg_param
*
@@ -991,9 +1008,8 @@ static u32 get_async_notif_value(optee_invoke_fn *invoke_fn, bool *value_valid,
return res.a1;
}

-static irqreturn_t notif_irq_handler(int irq, void *dev_id)
+static irqreturn_t irq_handler(struct optee *optee)
{
- struct optee *optee = dev_id;
bool do_bottom_half = false;
bool value_valid;
bool value_pending;
@@ -1016,6 +1032,13 @@ static irqreturn_t notif_irq_handler(int irq, void *dev_id)
return IRQ_HANDLED;
}

+static irqreturn_t notif_irq_handler(int irq, void *dev_id)
+{
+ struct optee *optee = dev_id;
+
+ return irq_handler(optee);
+}
+
static irqreturn_t notif_irq_thread_fn(int irq, void *dev_id)
{
struct optee *optee = dev_id;
@@ -1025,7 +1048,7 @@ static irqreturn_t notif_irq_thread_fn(int irq, void *dev_id)
return IRQ_HANDLED;
}

-static int optee_smc_notif_init_irq(struct optee *optee, u_int irq)
+static int init_irq(struct optee *optee, u_int irq)
{
int rc;

@@ -1040,12 +1063,106 @@ static int optee_smc_notif_init_irq(struct optee *optee, u_int irq)
return 0;
}

+static irqreturn_t notif_pcpu_irq_handler(int irq, void *dev_id)
+{
+ struct optee_pcpu __percpu *pcpu = (struct optee_pcpu *)dev_id;
+ struct optee *optee = pcpu->optee;
+
+ if (irq_handler(optee) == IRQ_WAKE_THREAD)
+ queue_work(optee->smc.notif_pcpu_wq,
+ &optee->smc.notif_pcpu_work);
+
+ return IRQ_HANDLED;
+}
+
+static void notif_pcpu_irq_work_fn(struct work_struct *work)
+{
+ struct optee_smc *optee_smc = container_of(work, struct optee_smc,
+ notif_pcpu_work);
+ struct optee *optee = container_of(optee_smc, struct optee, smc);
+
+ optee_smc_do_bottom_half(optee->ctx);
+}
+
+static int init_pcpu_irq(struct optee *optee, u_int irq)
+{
+ struct optee_pcpu __percpu *optee_pcpu;
+ int cpu, rc;
+
+ optee_pcpu = alloc_percpu(struct optee_pcpu);
+ if (!optee_pcpu)
+ return -ENOMEM;
+
+ for_each_present_cpu(cpu) {
+ struct optee_pcpu __percpu *p = per_cpu_ptr(optee_pcpu, cpu);
+
+ p->optee = optee;
+ }
+
+ rc = request_percpu_irq(irq, notif_pcpu_irq_handler,
+ "optee_pcpu_notification", optee_pcpu);
+ if (rc)
+ goto err_free_pcpu;
+
+ INIT_WORK(&optee->smc.notif_pcpu_work, notif_pcpu_irq_work_fn);
+ optee->smc.notif_pcpu_wq = create_workqueue("optee_pcpu_notification");
+ if (!optee->smc.notif_pcpu_wq) {
+ rc = -EINVAL;
+ goto err_free_pcpu_irq;
+ }
+
+ optee->smc.optee_pcpu = optee_pcpu;
+ optee->smc.notif_irq = irq;
+
+ pcpu_irq_num = irq;
+ rc = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "optee/pcpu-notif:starting",
+ optee_cpuhp_enable_pcpu_irq,
+ optee_cpuhp_disable_pcpu_irq);
+ if (!rc)
+ rc = -EINVAL;
+ if (rc < 0)
+ goto err_free_pcpu_irq;
+
+ optee->smc.notif_cpuhp_state = rc;
+
+ return 0;
+
+err_free_pcpu_irq:
+ free_percpu_irq(irq, optee_pcpu);
+err_free_pcpu:
+ free_percpu(optee_pcpu);
+
+ return rc;
+}
+
+static int optee_smc_notif_init_irq(struct optee *optee, u_int irq)
+{
+ if (irq_is_percpu_devid(irq))
+ return init_pcpu_irq(optee, irq);
+ else
+ return init_irq(optee, irq);
+}
+
+static void uninit_pcpu_irq(struct optee *optee)
+{
+ cpuhp_remove_state(optee->smc.notif_cpuhp_state);
+
+ destroy_workqueue(optee->smc.notif_pcpu_wq);
+
+ free_percpu_irq(optee->smc.notif_irq, optee->smc.optee_pcpu);
+ free_percpu(optee->smc.optee_pcpu);
+}
+
static void optee_smc_notif_uninit_irq(struct optee *optee)
{
if (optee->smc.sec_caps & OPTEE_SMC_SEC_CAP_ASYNC_NOTIF) {
optee_smc_stop_async_notif(optee->ctx);
if (optee->smc.notif_irq) {
- free_irq(optee->smc.notif_irq, optee);
+ if (irq_is_percpu_devid(optee->smc.notif_irq))
+ uninit_pcpu_irq(optee);
+ else
+ free_irq(optee->smc.notif_irq, optee);
+
irq_dispose_mapping(optee->smc.notif_irq);
}
}
--
2.25.1


2023-03-18 17:04:00

by kernel test robot

[permalink] [raw]
Subject: Re: [PATCH v5 2/2] optee: add per cpu asynchronous notification

Hi Etienne,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on soc/for-next]
[also build test WARNING on robh/for-next krzk-dt/for-next linus/master v6.3-rc2 next-20230317]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url: https://github.com/intel-lab-lkp/linux/commits/Etienne-Carriere/optee-add-per-cpu-asynchronous-notification/20230318-013004
base: https://git.kernel.org/pub/scm/linux/kernel/git/soc/soc.git for-next
patch link: https://lore.kernel.org/r/20230317172859.989650-2-etienne.carriere%40linaro.org
patch subject: [PATCH v5 2/2] optee: add per cpu asynchronous notification
config: arm64-randconfig-s053-20230312 (https://download.01.org/0day-ci/archive/20230319/[email protected]/config)
compiler: aarch64-linux-gcc (GCC) 12.1.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# apt-get install sparse
# sparse version: v0.6.4-39-gce1a6720-dirty
# https://github.com/intel-lab-lkp/linux/commit/58552ace9e840f772ba5cb7d4694e332c74802e8
git remote add linux-review https://github.com/intel-lab-lkp/linux
git fetch --no-tags linux-review Etienne-Carriere/optee-add-per-cpu-asynchronous-notification/20230318-013004
git checkout 58552ace9e840f772ba5cb7d4694e332c74802e8
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-12.1.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' O=build_dir ARCH=arm64 olddefconfig
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-12.1.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' O=build_dir ARCH=arm64 SHELL=/bin/bash drivers/firmware/smccc/ drivers/tee/optee/

If you fix the issue, kindly add following tag where applicable
| Reported-by: kernel test robot <[email protected]>
| Link: https://lore.kernel.org/oe-kbuild-all/[email protected]/

sparse warnings: (new ones prefixed by >>)
>> drivers/tee/optee/smc_abi.c:1068:45: sparse: sparse: incorrect type in initializer (different address spaces) @@ expected struct optee_pcpu [noderef] __percpu *pcpu @@ got struct optee_pcpu * @@
drivers/tee/optee/smc_abi.c:1068:45: sparse: expected struct optee_pcpu [noderef] __percpu *pcpu
drivers/tee/optee/smc_abi.c:1068:45: sparse: got struct optee_pcpu *
>> drivers/tee/optee/smc_abi.c:1097:49: sparse: sparse: incorrect type in initializer (different address spaces) @@ expected struct optee_pcpu [noderef] __percpu *p @@ got struct optee_pcpu * @@
drivers/tee/optee/smc_abi.c:1097:49: sparse: expected struct optee_pcpu [noderef] __percpu *p
drivers/tee/optee/smc_abi.c:1097:49: sparse: got struct optee_pcpu *
>> drivers/tee/optee/smc_abi.c:1069:31: sparse: sparse: dereference of noderef expression
drivers/tee/optee/smc_abi.c:1099:17: sparse: sparse: dereference of noderef expression

vim +1068 drivers/tee/optee/smc_abi.c

1065
1066 static irqreturn_t notif_pcpu_irq_handler(int irq, void *dev_id)
1067 {
> 1068 struct optee_pcpu __percpu *pcpu = (struct optee_pcpu *)dev_id;
> 1069 struct optee *optee = pcpu->optee;
1070
1071 if (irq_handler(optee) == IRQ_WAKE_THREAD)
1072 queue_work(optee->smc.notif_pcpu_wq,
1073 &optee->smc.notif_pcpu_work);
1074
1075 return IRQ_HANDLED;
1076 }
1077
1078 static void notif_pcpu_irq_work_fn(struct work_struct *work)
1079 {
1080 struct optee_smc *optee_smc = container_of(work, struct optee_smc,
1081 notif_pcpu_work);
1082 struct optee *optee = container_of(optee_smc, struct optee, smc);
1083
1084 optee_smc_do_bottom_half(optee->ctx);
1085 }
1086
1087 static int init_pcpu_irq(struct optee *optee, u_int irq)
1088 {
1089 struct optee_pcpu __percpu *optee_pcpu;
1090 int cpu, rc;
1091
1092 optee_pcpu = alloc_percpu(struct optee_pcpu);
1093 if (!optee_pcpu)
1094 return -ENOMEM;
1095
1096 for_each_present_cpu(cpu) {
> 1097 struct optee_pcpu __percpu *p = per_cpu_ptr(optee_pcpu, cpu);
1098
1099 p->optee = optee;
1100 }
1101
1102 rc = request_percpu_irq(irq, notif_pcpu_irq_handler,
1103 "optee_pcpu_notification", optee_pcpu);
1104 if (rc)
1105 goto err_free_pcpu;
1106
1107 INIT_WORK(&optee->smc.notif_pcpu_work, notif_pcpu_irq_work_fn);
1108 optee->smc.notif_pcpu_wq = create_workqueue("optee_pcpu_notification");
1109 if (!optee->smc.notif_pcpu_wq) {
1110 rc = -EINVAL;
1111 goto err_free_pcpu_irq;
1112 }
1113
1114 optee->smc.optee_pcpu = optee_pcpu;
1115 optee->smc.notif_irq = irq;
1116
1117 pcpu_irq_num = irq;
1118 rc = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "optee/pcpu-notif:starting",
1119 optee_cpuhp_enable_pcpu_irq,
1120 optee_cpuhp_disable_pcpu_irq);
1121 if (!rc)
1122 rc = -EINVAL;
1123 if (rc < 0)
1124 goto err_free_pcpu_irq;
1125
1126 optee->smc.notif_cpuhp_state = rc;
1127
1128 return 0;
1129
1130 err_free_pcpu_irq:
1131 free_percpu_irq(irq, optee_pcpu);
1132 err_free_pcpu:
1133 free_percpu(optee_pcpu);
1134
1135 return rc;
1136 }
1137

--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests

2023-03-19 12:17:23

by Krzysztof Kozlowski

[permalink] [raw]
Subject: Re: [PATCH v5 1/2] dt-bindings: optee driver interrupt can be a per-cpu interrupt

On 17/03/2023 18:28, Etienne Carriere wrote:
> Explicit in optee firmware device tree bindings that the interrupt
> used by optee driver for async notification can be a peripheral
> interrupt or a per-cpu interrupt.
>
> Signed-off-by: Etienne Carriere <[email protected]>
> ---
> No change since v4


Acked-by: Krzysztof Kozlowski <[email protected]>

Best regards,
Krzysztof