Linux kernel and device drivers for NXP i.MX platforms
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Nas Chung" <nas.chung@chipsnmedia.com>
Cc: devicetree@vger.kernel.org, imx@lists.linux.dev, robh@kernel.org,
	conor+dt@kernel.org, Frank.Li@kernel.org, media-ci@linuxtv.org
Subject: Re: [PATCH v8 02/11] media: chips-media: wave6: Add Wave6 control driver
Date: Thu, 19 Nov 2026 09:26:25 +0000	[thread overview]
Message-ID: <sashiko-outbox-222830@kernel.org> (raw)
In-Reply-To: <6b00047b748267a2ff4e18ccbfd573ea92166d34.1795075580.git.nas.chung@chipsnmedia.com>

Thank you for your contribution! Sashiko AI review found 8 potential
issues to consider:
- [High] media: wave6: vpu->lock held across core device register and
  unregister
- [High] media: wave6: data race on work_buffers_avail in
  wave6_vpu_require_work_buffer()
- [Medium] media: wave6: stuck PREPARE state on work buf error in
  wave6_vpu_init_vpu()
- [Medium] media: wave6: incompatible phys_addr_t pointer in
  wave6_vpu_probe()
- [Medium] media: wave6: wave6_vpu_probe() succeeds without required
  memory-region
- [Medium] media: wave6: wrong product code value for WAVE677_CODE
- [Low] media: wave6: off-by-one of_irq_get() error check in
  wave6_vpu_create_cores()
- [Low] media: wave6: mixed guard(mutex) and goto cleanup in
  wave6_vpu_get()
--

commit 600333b152222fe3a820d242ccd857c34cf7d58c
Author: Nas Chung <nas.chung@chipsnmedia.com>

media: chips-media: wave6: Add Wave6 control driver

This commit adds the control driver for the Chips&Media Wave6 video codec IP
to manage firmware loading, shared memory resources, and child core devices.

> diff --git a/drivers/media/platform/chips-media/wave6/wave6-vpu.c b/drivers/media/platform/chips-media/wave6/wave6-vpu.c
> new file mode 100644
> index 0000000000000..9160ce53062ce
> --- /dev/null
> +++ b/drivers/media/platform/chips-media/wave6/wave6-vpu.c
[ ... ]
> +static int wave6_vpu_init_vpu(struct wave6_vpu_device *vpu,
> +			      void __iomem *core_base)
> +{
> +	int ret;
> +
> +	lockdep_assert_held(&vpu->lock);
> +
> +	/* try init directly as firmware is running */
> +	if (wave6_vdi_readl(core_base, W6_VPU_VCPU_CUR_PC))
> +		goto init_done;
> +
> +	wave6_vpu_set_state(vpu, WAVE6_VPU_STATE_PREPARE);
> +
> +	wave6_vpu_remap_code_buf(vpu);
> +
> +	wave6_vdi_writel(core_base, W6_VPU_BUSY_STATUS, BUSY_STATUS_SET);
> +	wave6_vdi_writel(core_base, W6_CMD_INIT_VPU_SEC_AXI_BASE_CORE0,
> +			 vpu->sram_buf.dma_addr);
> +	wave6_vdi_writel(core_base, W6_CMD_INIT_VPU_SEC_AXI_SIZE_CORE0,
> +			 vpu->sram_buf.size);
> +	wave6_vdi_writel(vpu->reg_base, W6_COMMAND_GB, W6_CMD_INIT_VPU);
> +	wave6_vdi_writel(vpu->reg_base, W6_VPU_REMAP_CORE_START_GB,
> +			 REMAP_CORE_START_ON);
> +
> +	ret = wave6_vpu_wait_busy(core_base);
> +	if (ret) {
> +		dev_err(vpu->dev, "init vpu timeout\n");
> +		wave6_vpu_set_state(vpu, WAVE6_VPU_STATE_OFF);
> +		return -EINVAL;
> +	}
> +
> +	ret = wave6_vpu_check_result(core_base);
> +	if (ret) {
> +		dev_err(vpu->dev, "init vpu fail, reason 0x%x\n", ret);
> +		wave6_vpu_set_state(vpu, WAVE6_VPU_STATE_OFF);
> +		return -EIO;
> +	}
> +
> +init_done:
> +	ret = wave6_vpu_init_work_buf(vpu, core_base);
> +	if (ret)
> +		return ret;
> +
> +	wave6_vpu_set_state(vpu, WAVE6_VPU_STATE_ON);
> +
> +	return 0;
> +}

[Severity: Medium]
Does returning on wave6_vpu_init_work_buf() failure leave vpu->state stuck
in WAVE6_VPU_STATE_PREPARE?

Unlike the timeout and error paths above init_done, this path does not
reset vpu->state to WAVE6_VPU_STATE_OFF.  wave6_vpu_get() then powers down
the device and drops core_count to 0, and every later wave6_vpu_get() ->
wave6_vpu_try_boot() sees vpu->state == WAVE6_VPU_STATE_PREPARE and returns
0 without booting the VPU or initializing work buffers.

[ ... ]

> +int wave6_vpu_get(struct wave6_vpu_device *vpu, void __iomem *core_base)
> +{
> +	int ret;
> +
> +	guard(mutex)(&vpu->lock);
> +
> +	/* Only the first core executes boot; others return */
> +	if (atomic_inc_return(&vpu->core_count) > 1)
> +		return 0;
> +
> +	ret = pm_runtime_resume_and_get(vpu->dev);
> +	if (ret)
> +		goto error_pm;
> +
> +	ret = wave6_vpu_try_boot(vpu, core_base);
> +	if (ret)
> +		goto error_boot;
> +
> +	return 0;
> +
> +error_boot:
> +	pm_runtime_put_sync(vpu->dev);
> +error_pm:
> +	atomic_dec(&vpu->core_count);
> +
> +	return ret;
> +}

[Severity: Low]
This isn't a bug, but wave6_vpu_get() mixes guard(mutex) with goto-based
error unwinding in the same function, contrary to the Cleanup API
conventions in include/linux/cleanup.h.

[ ... ]

> +void wave6_vpu_require_work_buffer(struct wave6_vpu_device *vpu,
> +				   void __iomem *core_base)
> +{
> +	struct vpu_buf *vb;
> +	u32 size;
> +
> +	size = wave6_vdi_readl(core_base, W6_CMD_SET_WORK_BUF_SIZE);
> +	if (!size)
> +		return;
> +
> +	if (WARN_ON(size > W637DEC_WORKBUF_SIZE_FOR_CQ))
> +		goto exit;
> +
> +	if (WARN_ON(vpu->work_buffers_avail == 0))
> +		goto exit;
> +
> +	vpu->work_buffers_avail--;
> +	vb = &vpu->work_buffers[vpu->work_buffers_avail];
> +
> +	wave6_vdi_writel(core_base, W6_CMD_SET_WORK_BUF_ADDR, vb->daddr);
> +
> +exit:
> +	wave6_vdi_writel(core_base, W6_CMD_SET_WORK_BUF_SIZE, SET_WORK_BUF_SIZE_ACK);
> +}

[Severity: High]
Can concurrent calls to wave6_vpu_require_work_buffer() across VPU cores
race on vpu->work_buffers_avail?

In commit "media: chips-media: wave6: Add Wave6 core driver", each child
core device calls wave6_vpu_require_work_buffer() directly from its
wave6_vpu_core_irq() hardirq handler without holding any shared lock:

CPU0 (<hardirq> core 0)           CPU1 (<hardirq> core 1)
wave6_vpu_core_irq()              wave6_vpu_core_irq()
  wave6_vpu_require_work_buffer()   wave6_vpu_require_work_buffer()
    if (work_buffers_avail == 0)      if (work_buffers_avail == 0)
    work_buffers_avail--;             work_buffers_avail--;
    vb = &work_buffers[avail];        vb = &work_buffers[avail];
                                        <-- duplicate buffer or underflow

Two cores can be assigned the same work_buffers[] entry, or when
vpu->work_buffers_avail is 1, both cores can pass the zero check and
underflow the u32 counter to 0xffffffff, reading vpu->work_buffers[] out of
bounds.

Also, when jumping to exit on error, should W6_CMD_SET_WORK_BUF_ADDR be
cleared before writing SET_WORK_BUF_SIZE_ACK so the firmware does not
consume a stale address?

> +static int wave6_vpu_create_cores(struct wave6_vpu_device *vpu)
> +{
> +	struct device_node *child;
> +	int num_cores = 0;
> +
> +	for_each_available_child_of_node(vpu->dev->of_node, child) {
> +		struct platform_device_info info = {};
> +		struct platform_device *pdev;
> +		struct resource res[2];
> +		int irq;
> +
> +		if (num_cores >= W6_VPU_MAX_NUM_CORE) {
> +			of_node_put(child);
> +			break;
> +		}
> +
> +		if (of_address_to_resource(child, 0, &res[0])) {
> +			dev_warn(vpu->dev, "%pOF: missing reg property\n", child);
> +			continue;
> +		}
> +
> +		irq = of_irq_get(child, 0);
> +		if (irq < 0) {
> +			dev_warn(vpu->dev, "%pOF: missing interrupts property\n", child);
> +			continue;
> +		}
> +		res[1] = DEFINE_RES_IRQ(irq);

[Severity: Low]
Should the error check for of_irq_get() be "if (irq <= 0)"?

of_irq_get() returns 0 when IRQ mapping fails, which would register the
child platform device with an invalid IRQ 0 resource.

[ ... ]

> +static void wave6_vpu_release(struct wave6_vpu_device *vpu)
> +{
> +	guard(mutex)(&vpu->lock);
> +
> +	wave6_vpu_destroy_cores(vpu);
> +	wave6_vpu_free_work_buffers(vpu);
[ ... ]
> +}
> +
> +static void wave6_vpu_init_firmware(struct wave6_vpu_device *vpu,
> +				    const struct firmware *fw)
> +{
> +	guard(mutex)(&vpu->lock);
> +
> +	if (!fw || !fw->data) {
> +		dev_err(vpu->dev, "No firmware.\n");
> +		return;
> +	}
> +
> +	if (fw->size + W6_EXTRA_CODE_BUF_SIZE > wave6_vpu_get_code_buf_size(vpu)) {
> +		dev_err(vpu->dev, "firmware size (%zd > %zd) is too big\n",
> +			fw->size, vpu->code_buf.size);
> +		return;
> +	}
> +
> +	memcpy(vpu->code_buf.vaddr, fw->data, fw->size);
> +
> +	if (!wave6_vpu_create_cores(vpu))
> +		dev_err(vpu->dev, "Failed to create VPU cores\n");
> +}

[Severity: High]
Does holding vpu->lock across wave6_vpu_create_cores() and
wave6_vpu_destroy_cores() invert the lock hierarchy against child core
device locks and runtime PM transitions?

wave6_vpu_create_cores() -> platform_device_register_full() and
wave6_vpu_destroy_cores() -> platform_device_unregister() acquire the child
core device's device_lock (core_dev->mutex) while vpu->lock is held, and
platform_device_unregister() -> __device_release_driver() also runs
pm_runtime_get_sync(core_dev) and pm_runtime_put_sync(core_dev).  In commit
"media: chips-media: wave6: Add Wave6 core driver", the child's runtime PM
and system sleep callbacks call wave6_vpu_get() and wave6_vpu_put(), which
acquire vpu->lock under core_dev->mutex or during a core_dev runtime PM
transition:

CPU0                              CPU1
----                              ----
lock(&vpu->lock)
                                  lock(&core_dev->mutex)
lock(&core_dev->mutex)
                                  lock(&vpu->lock)

 *** DEADLOCK ***

CPU0: wave6_vpu_init_firmware() -> wave6_vpu_create_cores() ->
      platform_device_register_full()
CPU1: device_suspend() -> wave6_vpu_core_suspend() ->
      pm_runtime_force_suspend() -> wave6_vpu_core_runtime_suspend() ->
      wave6_vpu_put()

During teardown, if core_dev is in RPM_RESUMING or RPM_SUSPENDING waiting
for vpu->lock in wave6_vpu_get() or wave6_vpu_put() when
wave6_vpu_release() runs, __device_release_driver()'s
pm_runtime_get_sync(core_dev) will also wait on core_dev->power.wait_queue
while holding vpu->lock.

[ ... ]

> +	np = of_parse_phandle(pdev->dev.of_node, "memory-region", 0);
> +	if (np) {
> +		struct resource mem;
> +
> +		ret = of_address_to_resource(np, 0, &mem);
> +		of_node_put(np);
> +		if (!ret) {
> +			vpu->code_buf.phys_addr = mem.start;
> +			vpu->code_buf.size = resource_size(&mem);
> +			ret = wave6_vpu_init_code_buf(vpu);
> +			if (ret)
> +				goto error_detach_pm_domains;
> +		} else {
> +			dev_warn(&pdev->dev, "memory-region is not available.\n");
> +		}
> +	}

[Severity: Medium]
Should wave6_vpu_probe() fail and jump to error_detach_pm_domains when
"memory-region" is missing or of_address_to_resource() fails?

Without it, vpu->code_buf.size remains 0 while wave6_vpu_probe() continues,
allocates 48 MiB of coherent DMA work buffers and SRAM, and returns 0, only
for wave6_vpu_init_firmware() to unconditionally fail its size check against
wave6_vpu_get_code_buf_size(vpu) and leave those resources allocated without
creating any core devices.

> +	vpu->sram_pool = of_gen_pool_get(pdev->dev.of_node, "sram", 0);
> +	if (vpu->sram_pool) {
> +		vpu->sram_buf.size = vpu->res->sram_size;
> +		vpu->sram_buf.vaddr = gen_pool_dma_alloc(vpu->sram_pool,
> +							 vpu->sram_buf.size,
> +							 &vpu->sram_buf.phys_addr);

[Severity: Medium]
Does passing &vpu->sram_buf.phys_addr to gen_pool_dma_alloc() fail to build
when phys_addr_t and dma_addr_t have different widths?

struct wave6_vpu_dma_buf declares phys_addr as phys_addr_t, whereas
gen_pool_dma_alloc() takes a dma_addr_t * third argument, which triggers
-Werror=incompatible-pointer-types on configs where
CONFIG_PHYS_ADDR_T_64BIT and CONFIG_ARCH_DMA_ADDR_T_64BIT differ (exposed by
COMPILE_TEST).

> diff --git a/drivers/media/platform/chips-media/wave6/wave6-vpuconfig.h b/drivers/media/platform/chips-media/wave6/wave6-vpuconfig.h
> new file mode 100644
> index 0000000000000..1d4e016c2a582
> --- /dev/null
> +++ b/drivers/media/platform/chips-media/wave6/wave6-vpuconfig.h
> @@ -0,0 +1,37 @@
[ ... ]
> +#define WAVE617_CODE	0x6170
> +#define WAVE627_CODE	0x6270
> +#define WAVE633_CODE	0x6330
> +#define WAVE637_CODE	0x6370
> +#define WAVE663_CODE	0x6630
> +#define WAVE677_CODE	0x6670

[Severity: Medium]
Should WAVE677_CODE be 0x6770 instead of 0x6670?

All other Wave6 product codes encode the three-digit model number in bits
[15:4], so wave6_is_product_w_series() in wave6_vpu_try_boot() would reject
a Wave677 reporting 0x6770 with -EINVAL.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/cover.1795075577.git.nas.chung@chipsnmedia.com?part=2

  reply	other threads:[~2026-11-19  9:26 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-19  8:59 [PATCH v8 00/11] Add support for Wave6 video codec driver Nas Chung
2026-11-19  8:59 ` [PATCH v8 01/11] dt-bindings: media: nxp: Add Wave6 video codec device Nas Chung
2026-11-19  8:59 ` [PATCH v8 02/11] media: chips-media: wave6: Add Wave6 control driver Nas Chung
2026-11-19  9:26   ` sashiko-bot [this message]
2026-11-19  8:59 ` [PATCH v8 03/11] media: chips-media: wave6: Add Wave6 core driver Nas Chung
2026-11-19  8:59 ` [PATCH v8 04/11] media: chips-media: wave6: Add Wave6 decoder firmware interface Nas Chung
2026-11-19  9:22   ` sashiko-bot
2026-11-19  8:59 ` [PATCH v8 05/11] media: chips-media: wave6: Add Wave6 m2m decoder Nas Chung
2026-11-19  9:31   ` sashiko-bot
2026-11-19  9:00 ` [PATCH v8 06/11] media: chips-media: wave6: Add Wave6 encoder firmware interface Nas Chung
2026-11-19  9:10   ` sashiko-bot
2026-11-19  9:00 ` [PATCH v8 07/11] media: chips-media: wave6: Add Wave6 m2m encoder Nas Chung
2026-11-19  9:00 ` [PATCH v8 08/11] media: chips-media: wave6: Add trace and debugfs Nas Chung
2026-11-19  9:14   ` sashiko-bot
2026-11-19  9:00 ` [PATCH v8 09/11] media: chips-media: wave6: Add performance profiling Nas Chung
2026-11-19  9:20   ` sashiko-bot
2026-11-19  9:00 ` [PATCH v8 10/11] media: chips-media: wave6: Add Wave6 thermal cooling device Nas Chung
2026-11-19  9:21   ` sashiko-bot
2026-11-19  9:00 ` [PATCH v8 11/11] arm64: dts: freescale: imx95: Add video codec node Nas Chung

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-222830@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=Frank.Li@kernel.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=imx@lists.linux.dev \
    --cc=media-ci@linuxtv.org \
    --cc=nas.chung@chipsnmedia.com \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox