On Fri, Jan 09, 2026 at 11:07:40PM +0800, Aristo Chen wrote:
Hi Sumit
Sumit Garg sumit.garg@kernel.org 於 2026年1月9日週五 下午7:50寫道：
On Thu, Jan 08, 2026 at 02:45:09PM +0800, Aristo Chen wrote:
Collect OP-TEE OS revision from secure world for both SMC and FF-A ABIs, store it in the OP-TEE driver, and expose it through the generic get_tee_revision() callback.
Signed-off-by: Aristo Chen aristo.chen@canonical.com
drivers/tee/optee/core.c | 23 +++++++++++++ drivers/tee/optee/ffa_abi.c | 57 ++++++++++++++++++++++++------- drivers/tee/optee/optee_private.h | 19 +++++++++++ drivers/tee/optee/smc_abi.c | 15 ++++++-- 4 files changed, 99 insertions(+), 15 deletions(-)
diff --git a/drivers/tee/optee/core.c b/drivers/tee/optee/core.c index 5b62139714ce..2d807bc748bc 100644 --- a/drivers/tee/optee/core.c +++ b/drivers/tee/optee/core.c @@ -63,6 +63,29 @@ int optee_set_dma_mask(struct optee *optee, u_int pa_width) return dma_coerce_mask_and_coherent(&optee->teedev->dev, mask); }
+int optee_get_revision(struct tee_device *teedev, char *buf, size_t len) +{
struct optee *optee = tee_get_drvdata(teedev);u64 build_id;if (!optee)return -ENODEV;if (!buf || !len)return -EINVAL;build_id = optee->revision.os_build_id;if (build_id)scnprintf(buf, len, "%u.%u (%016llx)",optee->revision.os_major,optee->revision.os_minor,(unsigned long long)build_id);elsescnprintf(buf, len, "%u.%u", optee->revision.os_major,optee->revision.os_minor);return 0;+}
static void optee_bus_scan(struct work_struct *work) { WARN_ON(optee_enumerate_devices(PTA_CMD_GET_DEVICES_SUPP)); diff --git a/drivers/tee/optee/ffa_abi.c b/drivers/tee/optee/ffa_abi.c index bf8390789ecf..82dbed1c87e5 100644 --- a/drivers/tee/optee/ffa_abi.c +++ b/drivers/tee/optee/ffa_abi.c @@ -775,6 +775,42 @@ static int optee_ffa_reclaim_protmem(struct optee *optee,
- with a matching configuration.
*/
+static bool optee_ffa_get_os_revision(struct ffa_device *ffa_dev,
const struct ffa_ops *ops,struct optee_revision *revision,bool log)+{
const struct ffa_msg_ops *msg_ops = ops->msg_ops;struct ffa_send_direct_data data = {.data0 = OPTEE_FFA_GET_OS_VERSION,};int rc;msg_ops->mode_32bit_set(ffa_dev);rc = msg_ops->sync_send_receive(ffa_dev, &data);if (rc) {pr_err("Unexpected error %d\n", rc);return false;}if (revision) {revision->os_major = data.data0;revision->os_minor = data.data1;revision->os_build_id = data.data2;}if (log) {if (data.data2)pr_info("revision %lu.%lu (%08lx)",data.data0, data.data1, data.data2);elsepr_info("revision %lu.%lu", data.data0, data.data1);}return true;+}
static bool optee_ffa_api_is_compatible(struct ffa_device *ffa_dev, const struct ffa_ops *ops) { @@ -798,19 +834,8 @@ static bool optee_ffa_api_is_compatible(struct ffa_device *ffa_dev, return false; }
data = (struct ffa_send_direct_data){.data0 = OPTEE_FFA_GET_OS_VERSION,};rc = msg_ops->sync_send_receive(ffa_dev, &data);if (rc) {pr_err("Unexpected error %d\n", rc);
if (!optee_ffa_get_os_revision(ffa_dev, ops, NULL, true))Why do you need to invoke optee_ffa_get_os_revision() here?
I’m calling optee_ffa_get_os_revision() here to avoid duplicating the GET_OS_VERSION call/printing logic. The original code in optee_ffa_api_is_compatible() already did the OS version query and printed it; after factoring it out, the helper keeps that behavior in one place. If we don’t call it there, we’d either lose the existing dmesg log or re‑implement the same OS‑version query/print block again, which I’m trying to avoid.
If you’d rather avoid the extra call at this stage, maybe we can always print the log in the optee_ffa_get_os_revision() and remove the call from optee_ffa_api_is_compatible() entirely.
Yeah move the print to optee_ffa_get_os_revision() and drop the extra call.
-Sumit