Skip to contentSkip to conversation

scsi: ufs: core: Initialize hba->rpmbs list in ufshcd

PATCH Applied Ao Sun 4 participants 4 messages Jul 23 – Jul 29

Subsystems (2)
lore
Share & tools
lore

Ao Sun

author
#0 lore raw

PATCH 1/1 scsi: ufs: core: Initialize hba->rpmbs list in ufshcd

From: Ao Sun <[email protected]>

Initialize the hba->rpmbs list in ufshcd_alloc_host() to prevent NULL pointer dereference in the device teardown path if ufs_rpmb_probe() fails.

Fixes b06b8c421485 ("scsi: ufs: core: Add OP-TEE based RPMB driver for UFS devices")
---
 drivers/ufs/core/ufs-rpmb.c | 2 --
 drivers/ufs/core/ufshcd.c   | 1 +
 2 files changed, 1 insertion(+), 2 deletions(-)
drivers/ufs/core/ufs-rpmb.c +0 -2
index ffad049872b9..62120dc2e9da 100644
@@ -152,8 +152,6 @@ int ufs_rpmb_probe(struct ufs_hba *hba)
 		return -EINVAL;
 	}
 
-	INIT_LIST_HEAD(&hba->rpmbs);
-
 	struct rpmb_descr descr = {
 		.type = RPMB_TYPE_UFS,
 		.route_frames = ufs_rpmb_route_frames,
drivers/ufs/core/ufshcd.c +1 -0
index d3044a3089b5..60227069db06 100644
@@ -10988,6 +10988,7 @@ int ufshcd_alloc_host(struct device *dev, struct ufs_hba **hba_handle)
 	hba->nop_out_timeout = NOP_OUT_TIMEOUT;
 	ufshcd_set_sg_entry_size(hba, sizeof(struct ufshcd_sg_entry));
 	INIT_LIST_HEAD(&hba->clk_list_head);
+	INIT_LIST_HEAD(&hba->rpmbs);
 	spin_lock_init(&hba->outstanding_lock);
 
 	*hba_handle = hba;
Signature · 1 line
2.34.1
#1 lore raw
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] This problem wasn't introduced by this patch, but a circular dependency between the parent UFS RPMB device and its child RPMB device prevents both from being unregistered and freed.
- [High] This problem wasn't introduced by this patch, but the struct ufs_rpmb_dev embeds a struct device and is incorrectly allocated using devm_kzalloc(), violating the Linux device model and risking a Use-After-Free on unbind. --

--- Patch [1]: [PATCH] scsi: ufs: core: Initialize hba->rpmbs list in ufshcd --- commit b69c44ec9068da0584fce2f43d2e235ffbfe2675
Author: Ao Sun <[email protected]>

scsi: ufs: core: Initialize hba->rpmbs list in ufshcd

Initialize the hba->rpmbs list in ufshcd_alloc_host() to prevent NULL pointer dereference in the device teardown path if ufs_rpmb_probe() fails.

drivers/ufs/core/ufs-rpmb.c int ufs_rpmb_probe(struct ufs_hba *hba) · line 156

Quoted 10 lines
drivers/ufs/core/ufs-rpmb.c +0 -2
index ffad049872b9a..62120dc2e9da7 100644
@@ -152,8 +152,6 @@ int ufs_rpmb_probe(struct ufs_hba *hba)
 		return -EINVAL;
 	}
 
-	INIT_LIST_HEAD(&hba->rpmbs);
-

[Severity: High]
This isn't a bug introduced by this patch, but is it safe to allocate ufs_rpmb using devm_kzalloc() when it embeds a struct device?

The ufs_rpmb structure is tied to the host device's devres lifecycle:

ufs_rpmb_probe() {
        ...
        ufs_rpmb = devm_kzalloc(hba->dev, sizeof(*ufs_rpmb), GFP_KERNEL);
        ...
}

When the host controller unbinds, devres will free this memory. However, if a userspace process holds an open file descriptor to the sysfs attributes of ufs_rpmb->dev, the kobject reference count remains elevated.

When userspace later closes the file, the driver core will drop the final reference and invoke ufs_rpmb_device_release(), which will access the already-freed ufs_rpmb_dev structure.

Could this result in a use-after-free on unbind?

[Severity: High]
This isn't a bug introduced by this patch, but does the error path in ufs_rpmb_probe() fully clean up the registered RPMB devices?

If device_unregister() is called on ufs_rpmb->dev during probe error or unbind, the parent device is unregistered. However, rpmb_dev_register() creates a child device that holds a reference to this parent.

Because the child holds a reference, the parent's reference count will not drop to zero, preventing its release callback from ever running:

ufs_rpmb_device_release() {
        ...
        rpmb_dev_unregister(ufs_rpmb->rdev);
        ...
}

Since rpmb_dev_unregister() is called inside that release callback, neither device gets fully unregistered.

Does this create a circular dependency that leaks device structures and sysfs entries?

Quoted 3 lines
drivers/ufs/core/ufs-rpmb.c +0 -0
 	struct rpmb_descr descr = {
 		.type = RPMB_TYPE_UFS,
 		.route_frames = ufs_rpmb_route_frames,

[ ... ]

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Reviewed-by on v1 1/1
#2 lore raw
In reply to Ao Sun
Quoted 39 lines · Ao Sun

From: Ao Sun <[email protected]>

Initialize the hba->rpmbs list in ufshcd_alloc_host() to prevent NULL pointer dereference in the device teardown path if ufs_rpmb_probe() fails.

Fixes b06b8c421485 ("scsi: ufs: core: Add OP-TEE based RPMB driver for UFS

devices")

---
 drivers/ufs/core/ufs-rpmb.c | 2 --
 drivers/ufs/core/ufshcd.c   | 1 +

 2 files changed, 1 insertion(+), 2 deletions(-)
drivers/ufs/core/ufs-rpmb.c +0 -0
index ffad049872b9..62120dc2e9da 100644
@@ -152,8 +152,6 @@ int ufs_rpmb_probe(struct ufs_hba *hba)
                 return -EINVAL;
         }


-       INIT_LIST_HEAD(&hba->rpmbs);
-
        struct rpmb_descr descr = {
                .type = RPMB_TYPE_UFS, .route_frames = ufs_rpmb_route_frames,

drivers/ufs/core/ufshcd.c +1 -0
index d3044a3089b5..60227069db06 100644
@@ -10988,6 +10988,7 @@ int ufshcd_alloc_host(struct device *dev, struct
 ufs_hba **hba_handle)
         hba->nop_out_timeout = NOP_OUT_TIMEOUT;
         ufshcd_set_sg_entry_size(hba, sizeof(struct ufshcd_sg_entry));
         INIT_LIST_HEAD(&hba->clk_list_head);
+       INIT_LIST_HEAD(&hba->rpmbs);
         spin_lock_init(&hba->outstanding_lock);
Quoted 2 lines


        *hba_handle = hba;

Thanks for splitting this out, this is much easier to reason about.

For the two lifetime issues Sashiko reported (the container/rdev circular reference and the devm_kzalloc() ownership), both are real but pre-existing and independent of this fix. Let's keep this patch focused on the NULL deref and handle those in the teardown-order series, this one can go in on its own.

Kind regards,
Bean