Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <07c2de04-f4a7-71e2-ef41-64e87495ff42@collabora.com>
Date:   Thu, 31 Aug 2023 09:54:40 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.13.0
Subject: Re: [PATCH v5 02/10] media: videobuf2: Access vb2_queue bufs array
 through helper functions
To:     Hans Verkuil <hverkuil-cisco@xs4all.nl>, mchehab@kernel.org,
        tfiga@chromium.org, m.szyprowski@samsung.com, ming.qian@nxp.com,
        ezequiel@vanguardiasur.com.ar, p.zabel@pengutronix.de,
        gregkh@linuxfoundation.org, nicolas.dufresne@collabora.com
Cc:     linux-media@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-arm-kernel@lists.infradead.org,
        linux-mediatek@lists.infradead.org, linux-arm-msm@vger.kernel.org,
        linux-rockchip@lists.infradead.org, linux-staging@lists.linux.dev,
        kernel@collabora.com
References: <20230824092133.39510-1-benjamin.gaignard@collabora.com>
 <20230824092133.39510-3-benjamin.gaignard@collabora.com>
 <397c6779-6153-5df2-a9f7-708bb7fcf58d@xs4all.nl>
 <73e3e87b-78d6-13b2-8a20-09a75d5f23d0@collabora.com>
 <ac1a0848-8b95-0141-c590-f29d74e8c9da@xs4all.nl>
Content-Language: en-US
From:   Benjamin Gaignard <benjamin.gaignard@collabora.com>
In-Reply-To: <ac1a0848-8b95-0141-c590-f29d74e8c9da@xs4all.nl>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


Le 30/08/2023 à 18:36, Hans Verkuil a écrit :
> On 30/08/2023 18:24, Benjamin Gaignard wrote:
>> Le 30/08/2023 à 15:23, Hans Verkuil a écrit :
>>> On 24/08/2023 11:21, Benjamin Gaignard wrote:
>>>> The first step before changing how vb2 buffers are stored into queue
>>>> is to avoid direct access to bufs arrays.
>>>>
>>>> This patch adds 2 helpers functions to add and remove vb2 buffers
>>>> from a queue. With these 2 and vb2_get_buffer(), bufs field of
>>>> struct vb2_queue becomes like a private member of the structure.
>>>>
>>>> After each call to vb2_get_buffer() we need to be sure that we get
>>>> a valid pointer so check the return value of all of them.
>>>>
>>>> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
>>>>
>>>> # Conflicts:
>>>> #    drivers/media/common/videobuf2/videobuf2-core.c
>>>> ---
>>>>    .../media/common/videobuf2/videobuf2-core.c   | 203 ++++++++++++++----
>>>>    .../media/common/videobuf2/videobuf2-v4l2.c   |  28 ++-
>>>>    drivers/media/platform/amphion/vpu_dbg.c      |  22 +-
>>>>    .../platform/mediatek/jpeg/mtk_jpeg_core.c    |   6 +-
>>>>    .../vcodec/decoder/vdec/vdec_vp9_req_lat_if.c |   2 +-
>>>>    drivers/media/platform/st/sti/hva/hva-v4l2.c  |   4 +
>>>>    drivers/media/test-drivers/visl/visl-dec.c    |  28 ++-
>>>>    .../staging/media/atomisp/pci/atomisp_ioctl.c |   2 +-
>>>>    8 files changed, 230 insertions(+), 65 deletions(-)
>>>>
>>>> diff --git a/drivers/media/common/videobuf2/videobuf2-core.c b/drivers/media/common/videobuf2/videobuf2-core.c
>>>> index e06905533ef4..8aa13591b782 100644
>>>> --- a/drivers/media/common/videobuf2/videobuf2-core.c
>>>> +++ b/drivers/media/common/videobuf2/videobuf2-core.c
>>>> @@ -403,6 +403,37 @@ static void init_buffer_cache_hints(struct vb2_queue *q, struct vb2_buffer *vb)
>>>>            vb->skip_cache_sync_on_finish = 1;
>>>>    }
>>>>    +/**
>>>> + * vb2_queue_add_buffer() - add a buffer to a queue
>>>> + * @q:    pointer to &struct vb2_queue with videobuf2 queue.
>>>> + * @vb:    pointer to &struct vb2_buffer to be added to the queue.
>>>> + * @index: index where add vb2_buffer in the queue
>>>> + */
>>>> +static bool vb2_queue_add_buffer(struct vb2_queue *q, struct vb2_buffer *vb, int index)
>>>> +{
>>>> +    if (index < VB2_MAX_FRAME && !q->bufs[index]) {
>>>> +        q->bufs[index] = vb;
>>>> +        vb->index = index;
>>>> +        vb->vb2_queue = q;
>>>> +        return true;
>>>> +    }
>>>> +
>>>> +    return false;
>>>> +}
>>>> +
>>>> +/**
>>>> + * vb2_queue_remove_buffer() - remove a buffer from a queue
>>>> + * @q:    pointer to &struct vb2_queue with videobuf2 queue.
>>>> + * @vb:    pointer to &struct vb2_buffer to be removed from the queue.
>>>> + */
>>>> +static void vb2_queue_remove_buffer(struct vb2_queue *q, struct vb2_buffer *vb)
>>>> +{
>>>> +    if (vb->index < VB2_MAX_FRAME) {
>>>> +        q->bufs[vb->index] = NULL;
>>>> +        vb->vb2_queue = NULL;
>>>> +    }
>>>> +}
>>>> +
>>>>    /*
>>>>     * __vb2_queue_alloc() - allocate vb2 buffer structures and (for MMAP type)
>>>>     * video buffer memory for all buffers/planes on the queue and initializes the
>>>> @@ -431,9 +462,7 @@ static int __vb2_queue_alloc(struct vb2_queue *q, enum vb2_memory memory,
>>>>            }
>>>>              vb->state = VB2_BUF_STATE_DEQUEUED;
>>>> -        vb->vb2_queue = q;
>>>>            vb->num_planes = num_planes;
>>>> -        vb->index = q->num_buffers + buffer;
>>>>            vb->type = q->type;
>>>>            vb->memory = memory;
>>>>            init_buffer_cache_hints(q, vb);
>>>> @@ -443,7 +472,11 @@ static int __vb2_queue_alloc(struct vb2_queue *q, enum vb2_memory memory,
>>>>            }
>>>>            call_void_bufop(q, init_buffer, vb);
>>>>    -        q->bufs[vb->index] = vb;
>>>> +        if (!vb2_queue_add_buffer(q, vb, q->num_buffers + buffer)) {
>>>> +            dprintk(q, 1, "failed adding buffer %d to queue\n", buffer);
>>>> +            kfree(vb);
>>>> +            break;
>>>> +        }
>>>>              /* Allocate video buffer memory for the MMAP type */
>>>>            if (memory == VB2_MEMORY_MMAP) {
>>>> @@ -451,7 +484,7 @@ static int __vb2_queue_alloc(struct vb2_queue *q, enum vb2_memory memory,
>>>>                if (ret) {
>>>>                    dprintk(q, 1, "failed allocating memory for buffer %d\n",
>>>>                        buffer);
>>>> -                q->bufs[vb->index] = NULL;
>>>> +                vb2_queue_remove_buffer(q, vb);
>>>>                    kfree(vb);
>>>>                    break;
>>>>                }
>>>> @@ -466,7 +499,7 @@ static int __vb2_queue_alloc(struct vb2_queue *q, enum vb2_memory memory,
>>>>                    dprintk(q, 1, "buffer %d %p initialization failed\n",
>>>>                        buffer, vb);
>>>>                    __vb2_buf_mem_free(vb);
>>>> -                q->bufs[vb->index] = NULL;
>>>> +                vb2_queue_remove_buffer(q, vb);
>>>>                    kfree(vb);
>>>>                    break;
>>>>                }
>>>> @@ -489,7 +522,7 @@ static void __vb2_free_mem(struct vb2_queue *q, unsigned int buffers)
>>>>          for (buffer = q->num_buffers - buffers; buffer < q->num_buffers;
>>>>             ++buffer) {
>>>> -        vb = q->bufs[buffer];
>>>> +        vb = vb2_get_buffer(q, buffer);
>>>>            if (!vb)
>>>>                continue;
>>>>    @@ -517,7 +550,7 @@ static void __vb2_queue_free(struct vb2_queue *q, unsigned int buffers)
>>>>        /* Call driver-provided cleanup function for each buffer, if provided */
>>>>        for (buffer = q->num_buffers - buffers; buffer < q->num_buffers;
>>>>             ++buffer) {
>>>> -        struct vb2_buffer *vb = q->bufs[buffer];
>>>> +        struct vb2_buffer *vb = vb2_get_buffer(q, buffer);
>>>>              if (vb && vb->planes[0].mem_priv)
>>>>                call_void_vb_qop(vb, buf_cleanup, vb);
>>>> @@ -557,15 +590,20 @@ static void __vb2_queue_free(struct vb2_queue *q, unsigned int buffers)
>>>>            q->cnt_unprepare_streaming = 0;
>>>>        }
>>>>        for (buffer = 0; buffer < q->num_buffers; ++buffer) {
>>>> -        struct vb2_buffer *vb = q->bufs[buffer];
>>>> -        bool unbalanced = vb->cnt_mem_alloc != vb->cnt_mem_put ||
>>>> -                  vb->cnt_mem_prepare != vb->cnt_mem_finish ||
>>>> -                  vb->cnt_mem_get_userptr != vb->cnt_mem_put_userptr ||
>>>> -                  vb->cnt_mem_attach_dmabuf != vb->cnt_mem_detach_dmabuf ||
>>>> -                  vb->cnt_mem_map_dmabuf != vb->cnt_mem_unmap_dmabuf ||
>>>> -                  vb->cnt_buf_queue != vb->cnt_buf_done ||
>>>> -                  vb->cnt_buf_prepare != vb->cnt_buf_finish ||
>>>> -                  vb->cnt_buf_init != vb->cnt_buf_cleanup;
>>>> +        struct vb2_buffer *vb = vb2_get_buffer(q, buffer);
>>>> +        bool unbalanced;
>>>> +
>>>> +        if (!vb)
>>>> +            continue;
>>>> +
>>>> +        unbalanced = vb->cnt_mem_alloc != vb->cnt_mem_put ||
>>>> +                 vb->cnt_mem_prepare != vb->cnt_mem_finish ||
>>>> +                 vb->cnt_mem_get_userptr != vb->cnt_mem_put_userptr ||
>>>> +                 vb->cnt_mem_attach_dmabuf != vb->cnt_mem_detach_dmabuf ||
>>>> +                 vb->cnt_mem_map_dmabuf != vb->cnt_mem_unmap_dmabuf ||
>>>> +                 vb->cnt_buf_queue != vb->cnt_buf_done ||
>>>> +                 vb->cnt_buf_prepare != vb->cnt_buf_finish ||
>>>> +                 vb->cnt_buf_init != vb->cnt_buf_cleanup;
>>>>              if (unbalanced || debug) {
>>> I think we should drop the '|| debug' part. It is already annoying today to see these
>>> messages when the debug parameter is > 0, and now the number of buffers is still
>>> fairly small. But if we allow a lot more buffers, then this will really spam the
>>> kernel log.
>>>
>>> I think this should be dropped, and we only report unbalanced buffers.
>>>
>>> And another optimization is to only report the unbalanced counters. Right now
>>> it reports all counters, but it is again too much spamming of the kernel log.
>>>
>>> I think this change can be done as a separate patch before this patch.
>>> That way it can be picked up separately from the other changes in this series.
>>>
>>>>                pr_info("   counters for queue %p, buffer %d:%s\n",
>>>> @@ -597,8 +635,13 @@ static void __vb2_queue_free(struct vb2_queue *q, unsigned int buffers)
>>>>        /* Free vb2 buffers */
>>>>        for (buffer = q->num_buffers - buffers; buffer < q->num_buffers;
>>>>             ++buffer) {
>>>> -        kfree(q->bufs[buffer]);
>>>> -        q->bufs[buffer] = NULL;
>>>> +        struct vb2_buffer *vb = vb2_get_buffer(q, buffer);
>>>> +
>>>> +        if (!vb)
>>>> +            continue;
>>>> +
>>>> +        vb2_queue_remove_buffer(q, vb);
>>>> +        kfree(vb);
>>>>        }
>>>>          q->num_buffers -= buffers;
>>>> @@ -634,7 +677,12 @@ static bool __buffers_in_use(struct vb2_queue *q)
>>>>    {
>>>>        unsigned int buffer;
>>>>        for (buffer = 0; buffer < q->num_buffers; ++buffer) {
>>>> -        if (vb2_buffer_in_use(q, q->bufs[buffer]))
>>>> +        struct vb2_buffer *vb = vb2_get_buffer(q, buffer);
>>>> +
>>>> +        if (!vb)
>>>> +            continue;
>>>> +
>>>> +        if (vb2_buffer_in_use(q, vb))
>>>>                return true;
>>>>        }
>>>>        return false;
>>>> @@ -642,7 +690,10 @@ static bool __buffers_in_use(struct vb2_queue *q)
>>>>      void vb2_core_querybuf(struct vb2_queue *q, unsigned int index, void *pb)
>>>>    {
>>>> -    call_void_bufop(q, fill_user_buffer, q->bufs[index], pb);
>>>> +    struct vb2_buffer *vb = vb2_get_buffer(q, index);
>>>> +
>>>> +    if (vb)
>>>> +        call_void_bufop(q, fill_user_buffer, vb, pb);
>>> I think that rather than passing the index (that then has to be verified)
>>> it is better to pass the vb2_buffer pointer directly and leave it up to
>>> the caller to do the index verification.
>>>
>>> Another option is to drop this function altogether and let the called
>>> call the fill_user_buffer function. Either works for me.
>>>
>>>>    }
>>>>    EXPORT_SYMBOL_GPL(vb2_core_querybuf);
>>>>    @@ -1553,7 +1604,13 @@ int vb2_core_prepare_buf(struct vb2_queue *q, unsigned int index, void *pb)
>>> Here too it is better to pass the vb2_buffer pointer instead of an index.
>>>
>>> This function assumes that the index is valid, so the called actually does the
>>> validation. Passing the vb pointer instead of the index makes more sense
>>> in this new situation.
>>>
>>> This is also true for two other core functions: vb2_core_qbuf and vb2_core_expbuf.
>>>
>>>>        struct vb2_buffer *vb;
>>>>        int ret;
>>>>    -    vb = q->bufs[index];
>>>> +    vb = vb2_get_buffer(q, index);
>>>> +
>>>> +    if (!vb) {
>>>> +        dprintk(q, 1, "can't find the requested buffer\n");
>>>> +        return -EINVAL;
>>>> +    }
>>> Changing that avoids having to add this check, so it simplifies the code.
>>>
>>> I think that this change can be done in a separate patch before this one.
>>>
>>> It makes sense to apply that regardless of the remainder of this series.
>>>
>>>> +
>>>>        if (vb->state != VB2_BUF_STATE_DEQUEUED) {
>>>>            dprintk(q, 1, "invalid buffer state %s\n",
>>>>                vb2_state_name(vb->state));
>>>> @@ -1624,7 +1681,11 @@ static int vb2_start_streaming(struct vb2_queue *q)
>>>>             * correctly return them to vb2.
>>>>             */
>>>>            for (i = 0; i < q->num_buffers; ++i) {
>>>> -            vb = q->bufs[i];
>>>> +            vb = vb2_get_buffer(q, i);
>>>> +
>>>> +            if (!vb)
>>>> +                continue;
>>>> +
>>>>                if (vb->state == VB2_BUF_STATE_ACTIVE)
>>>>                    vb2_buffer_done(vb, VB2_BUF_STATE_QUEUED);
>>>>            }
>>>> @@ -1652,7 +1713,12 @@ int vb2_core_qbuf(struct vb2_queue *q, unsigned int index, void *pb,
>>>>            return -EIO;
>>>>        }
>>>>    -    vb = q->bufs[index];
>>>> +    vb = vb2_get_buffer(q, index);
>>>> +
>>>> +    if (!vb) {
>>>> +        dprintk(q, 1, "can't find the requested buffer\n");
>>>> +        return -EINVAL;
>>>> +    }
>>>>          if (!req && vb->state != VB2_BUF_STATE_IN_REQUEST &&
>>>>            q->requires_requests) {
>>>> @@ -2028,12 +2094,18 @@ static void __vb2_queue_cancel(struct vb2_queue *q)
>>>>         * to vb2 in stop_streaming().
>>>>         */
>>>>        if (WARN_ON(atomic_read(&q->owned_by_drv_count))) {
>>>> -        for (i = 0; i < q->num_buffers; ++i)
>>>> -            if (q->bufs[i]->state == VB2_BUF_STATE_ACTIVE) {
>>>> +        for (i = 0; i < q->num_buffers; ++i) {
>>>> +            struct vb2_buffer *vb = vb2_get_buffer(q, i);
>>>> +
>>>> +            if (!vb)
>>>> +                continue;
>>>> +
>>>> +            if (vb->state == VB2_BUF_STATE_ACTIVE) {
>>>>                    pr_warn("driver bug: stop_streaming operation is leaving buf %p in active state\n",
>>>> -                    q->bufs[i]);
>>>> -                vb2_buffer_done(q->bufs[i], VB2_BUF_STATE_ERROR);
>>>> +                    vb);
>>>> +                vb2_buffer_done(vb, VB2_BUF_STATE_ERROR);
>>>>                }
>>>> +        }
>>>>            /* Must be zero now */
>>>>            WARN_ON(atomic_read(&q->owned_by_drv_count));
>>>>        }
>>>> @@ -2067,9 +2139,14 @@ static void __vb2_queue_cancel(struct vb2_queue *q)
>>>>         * be changed, so we can't move the buf_finish() to __vb2_dqbuf().
>>>>         */
>>>>        for (i = 0; i < q->num_buffers; ++i) {
>>>> -        struct vb2_buffer *vb = q->bufs[i];
>>>> -        struct media_request *req = vb->req_obj.req;
>>>> +        struct vb2_buffer *vb;
>>>> +        struct media_request *req;
>>>> +
>>>> +        vb = vb2_get_buffer(q, i);
>>>> +        if (!vb)
>>>> +            continue;
>>>>    +        req = vb->req_obj.req;
>>>>            /*
>>>>             * If a request is associated with this buffer, then
>>>>             * call buf_request_cancel() to give the driver to complete()
>>>> @@ -2219,7 +2296,10 @@ static int __find_plane_by_offset(struct vb2_queue *q, unsigned long off,
>>>>        buffer = (off >> (PLANE_INDEX_SHIFT + PAGE_SHIFT)) & MAX_BUFFERS;
>>>>        plane = (off >> PAGE_SHIFT) & PLANE_INDEX_MASK;
>>>>    -    vb = q->bufs[buffer];
>>>> +    vb = vb2_get_buffer(q, buffer);
>>>> +    if (!vb)
>>>> +        return -EINVAL;
>>>> +
>>>>        if (vb->planes[plane].m.offset == off) {
>>>>            *_buffer = buffer;
>>>>            *_plane = plane;
>>>> @@ -2262,7 +2342,12 @@ int vb2_core_expbuf(struct vb2_queue *q, int *fd, unsigned int type,
>>>>            return -EINVAL;
>>>>        }
>>>>    -    vb = q->bufs[index];
>>>> +    vb = vb2_get_buffer(q, index);
>>>> +
>>>> +    if (!vb) {
>>>> +        dprintk(q, 1, "can't find the requested buffer\n");
>>>> +        return -EINVAL;
>>>> +    }
>>>>          if (plane >= vb->num_planes) {
>>>>            dprintk(q, 1, "buffer plane out of range\n");
>>>> @@ -2339,7 +2424,13 @@ int vb2_mmap(struct vb2_queue *q, struct vm_area_struct *vma)
>>>>        if (ret)
>>>>            goto unlock;
>>>>    -    vb = q->bufs[buffer];
>>>> +    vb = vb2_get_buffer(q, buffer);
>>>> +
>>>> +    if (!vb) {
>>>> +        dprintk(q, 1, "can't find the requested buffer\n");
>>>> +        ret = -EINVAL;
>>>> +        goto unlock;
>>>> +    }
>>>>          /*
>>>>         * MMAP requires page_aligned buffers.
>>>> @@ -2396,7 +2487,12 @@ unsigned long vb2_get_unmapped_area(struct vb2_queue *q,
>>>>        if (ret)
>>>>            goto unlock;
>>>>    -    vb = q->bufs[buffer];
>>>> +    vb = vb2_get_buffer(q, buffer);
>>>> +    if (!vb) {
>>>> +        dprintk(q, 1, "can't find the requested buffer\n");
>>>> +        ret = -EINVAL;
>>>> +        goto unlock;
>>>> +    }
>>>>          vaddr = vb2_plane_vaddr(vb, plane);
>>>>        mutex_unlock(&q->mmap_lock);
>>>> @@ -2625,6 +2721,7 @@ struct vb2_fileio_data {
>>>>    static int __vb2_init_fileio(struct vb2_queue *q, int read)
>>>>    {
>>>>        struct vb2_fileio_data *fileio;
>>>> +    struct vb2_buffer *vb;
>>>>        int i, ret;
>>>>        unsigned int count = 0;
>>>>    @@ -2679,7 +2776,13 @@ static int __vb2_init_fileio(struct vb2_queue *q, int read)
>>>>         * Check if plane_count is correct
>>>>         * (multiplane buffers are not supported).
>>>>         */
>>>> -    if (q->bufs[0]->num_planes != 1) {
>>>> +    vb = vb2_get_buffer(q, 0);
>>>> +    if (!vb) {
>>>> +        ret = -EBUSY;
>>>> +        goto err_reqbufs;
>>>> +    }
>>> This cannot happen. These fileio helper functions implement the read() support
>>> and all the buffer allocation happens here. Userspace can never add or delete
>>> buffers later, so there will never be holes. It is safe to assume that
>>> vb2_get_buffer(q, i) will always return a valid vb pointer for i in the range
>>> of 0 - q->num_buffers-1.
>>>
>>> Perhaps add a comment to that effect, but otherwise you can drop the checks.
>>>
>>>> +
>>>> +    if (vb->num_planes != 1) {
>>>>            ret = -EBUSY;
>>>>            goto err_reqbufs;
>>>>        }
>>>> @@ -2688,12 +2791,17 @@ static int __vb2_init_fileio(struct vb2_queue *q, int read)
>>>>         * Get kernel address of each buffer.
>>>>         */
>>>>        for (i = 0; i < q->num_buffers; i++) {
>>>> -        fileio->bufs[i].vaddr = vb2_plane_vaddr(q->bufs[i], 0);
>>>> +        vb = vb2_get_buffer(q, i);
>>>> +
>>>> +        if (!vb)
>>>> +            continue;
>>>> +
>>>> +        fileio->bufs[i].vaddr = vb2_plane_vaddr(vb, 0);
>>>>            if (fileio->bufs[i].vaddr == NULL) {
>>>>                ret = -EINVAL;
>>>>                goto err_reqbufs;
>>>>            }
>>>> -        fileio->bufs[i].size = vb2_plane_size(q->bufs[i], 0);
>>>> +        fileio->bufs[i].size = vb2_plane_size(vb, 0);
>>>>        }
>>>>          /*
>>>> @@ -2821,15 +2929,18 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
>>>>              fileio->cur_index = index;
>>>>            buf = &fileio->bufs[index];
>>>> -        b = q->bufs[index];
>>>> +        b = vb2_get_buffer(q, index);
>>>> +
>>>> +        if (!b)
>>>> +            return -EINVAL;
>>>>              /*
>>>>             * Get number of bytes filled by the driver
>>>>             */
>>>>            buf->pos = 0;
>>>>            buf->queued = 0;
>>>> -        buf->size = read ? vb2_get_plane_payload(q->bufs[index], 0)
>>>> -                 : vb2_plane_size(q->bufs[index], 0);
>>>> +        buf->size = read ? vb2_get_plane_payload(b, 0)
>>>> +                 : vb2_plane_size(b, 0);
>>>>            /* Compensate for data_offset on read in the multiplanar case. */
>>>>            if (is_multiplanar && read &&
>>>>                    b->planes[0].data_offset < buf->size) {
>>>> @@ -2872,8 +2983,12 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
>>>>         * Queue next buffer if required.
>>>>         */
>>>>        if (buf->pos == buf->size || (!read && fileio->write_immediately)) {
>>>> -        struct vb2_buffer *b = q->bufs[index];
>>>> +        struct vb2_buffer *b = vb2_get_buffer(q, index);
>>>>    +        if (!b) {
>>>> +            dprintk(q, 1, "can't find the requested buffer\n");
>>>> +            return -EINVAL;
>>>> +        }
>>>>            /*
>>>>             * Check if this is the last buffer to read.
>>>>             */
>>>> @@ -2899,7 +3014,7 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
>>>>             */
>>>>            buf->pos = 0;
>>>>            buf->queued = 1;
>>>> -        buf->size = vb2_plane_size(q->bufs[index], 0);
>>>> +        buf->size = vb2_plane_size(b, 0);
>>>>            fileio->q_count += 1;
>>>>            /*
>>>>             * If we are queuing up buffers for the first time, then
>>>> @@ -2970,7 +3085,9 @@ static int vb2_thread(void *data)
>>>>             * Call vb2_dqbuf to get buffer back.
>>>>             */
>>>>            if (prequeue) {
>>>> -            vb = q->bufs[index++];
>>>> +            vb = vb2_get_buffer(q, index++);
>>>> +            if (!vb)
>>>> +                continue;
>>>>                prequeue--;
>>>>            } else {
>>>>                call_void_qop(q, wait_finish, q);
>>>> @@ -2979,7 +3096,7 @@ static int vb2_thread(void *data)
>>>>                call_void_qop(q, wait_prepare, q);
>>>>                dprintk(q, 5, "file io: vb2_dqbuf result: %d\n", ret);
>>>>                if (!ret)
>>>> -                vb = q->bufs[index];
>>>> +                vb = vb2_get_buffer(q, index);
>>>>            }
>>>>            if (ret || threadio->stop)
>>>>                break;
>>>> diff --git a/drivers/media/common/videobuf2/videobuf2-v4l2.c b/drivers/media/common/videobuf2/videobuf2-v4l2.c
>>>> index c7a54d82a55e..724135d41f7f 100644
>>>> --- a/drivers/media/common/videobuf2/videobuf2-v4l2.c
>>>> +++ b/drivers/media/common/videobuf2/videobuf2-v4l2.c
>>>> @@ -383,8 +383,7 @@ static int vb2_queue_or_prepare_buf(struct vb2_queue *q, struct media_device *md
>>>>            return -EINVAL;
>>>>        }
>>>>    -    if (q->bufs[b->index] == NULL) {
>>>> -        /* Should never happen */
>>>> +    if (!vb2_get_buffer(q, b->index)) {
>>>>            dprintk(q, 1, "%s: buffer is NULL\n", opname);
>>> How about:
>>>
>>>             dprintk(q, 1, "%s: buffer %u was deleted\n", opname, b->index);
>>>
>>> although perhaps that change is more appropriate in patch 09/10?
>>>
>>> Regardless, once it is possible to delete buffers, then this message should be
>>> adjusted accordingly.
>>>
>>>>            return -EINVAL;
>>>>        }
>>>> @@ -394,7 +393,7 @@ static int vb2_queue_or_prepare_buf(struct vb2_queue *q, struct media_device *md
>>>>            return -EINVAL;
>>>>        }
>>>>    -    vb = q->bufs[b->index];
>>>> +    vb = vb2_get_buffer(q, b->index);
>>> This can be moved up to the 'if (!vb2_get_buffer(q, b->index)) {' check above.
>>> That avoids calling vb2_get_buffer twice.
>>>
>>>>        vbuf = to_vb2_v4l2_buffer(vb);
>>>>        ret = __verify_planes_array(vb, b);
>>>>        if (ret)
>>>> @@ -628,11 +627,18 @@ static const struct vb2_buf_ops v4l2_buf_ops = {
>>>>    struct vb2_buffer *vb2_find_buffer(struct vb2_queue *q, u64 timestamp)
>>>>    {
>>>>        unsigned int i;
>>>> +    struct vb2_buffer *vb2;
>>>>    -    for (i = 0; i < q->num_buffers; i++)
>>>> -        if (q->bufs[i]->copied_timestamp &&
>>>> -            q->bufs[i]->timestamp == timestamp)
>>>> -            return vb2_get_buffer(q, i);
>>> Perhaps add a comment here that this loop doesn't scale if there
>>> is a really large number of buffers and something more efficient
>>> will have to be found in that case.
>>>
>>>> +    for (i = 0; i < q->num_buffers; i++) {
>>>> +        vb2 = vb2_get_buffer(q, i);
>>>> +
>>>> +        if (!vb2)
>>>> +            continue;
>>>> +
>>>> +        if (vb2->copied_timestamp &&
>>>> +            vb2->timestamp == timestamp)
>>>> +            return vb2;
>>>> +    }
>>>>        return NULL;
>>>>    }
>>>>    EXPORT_SYMBOL_GPL(vb2_find_buffer);
>>>> @@ -664,7 +670,13 @@ int vb2_querybuf(struct vb2_queue *q, struct v4l2_buffer *b)
>>>>            dprintk(q, 1, "buffer index out of range\n");
>>>>            return -EINVAL;
>>>>        }
>>>> -    vb = q->bufs[b->index];
>>>> +    vb = vb2_get_buffer(q, b->index);
>>>> +
>>>> +    if (!vb) {
>>>> +        dprintk(q, 1, "can't find the requested buffer\n");
>>>> +        return -EINVAL;
>>>> +    }
>>>> +
>>>>        ret = __verify_planes_array(vb, b);
>>>>        if (!ret)
>>>>            vb2_core_querybuf(q, b->index, b);
>>>> diff --git a/drivers/media/platform/amphion/vpu_dbg.c b/drivers/media/platform/amphion/vpu_dbg.c
>>>> index 982c2c777484..a462d6fe4ea9 100644
>>>> --- a/drivers/media/platform/amphion/vpu_dbg.c
>>>> +++ b/drivers/media/platform/amphion/vpu_dbg.c
>>>> @@ -140,11 +140,18 @@ static int vpu_dbg_instance(struct seq_file *s, void *data)
>>>>          vq = v4l2_m2m_get_src_vq(inst->fh.m2m_ctx);
>>>>        for (i = 0; i < vq->num_buffers; i++) {
>>>> -        struct vb2_buffer *vb = vq->bufs[i];
>>>> -        struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
>>>> +        struct vb2_buffer *vb;
>>>> +        struct vb2_v4l2_buffer *vbuf;
>>>> +
>>>> +        vb = vb2_get_buffer(vq, i);
>>>> +        if (!vb)
>>>> +            continue;
>>>>              if (vb->state == VB2_BUF_STATE_DEQUEUED)
>>>>                continue;
>>>> +
>>>> +        vbuf = to_vb2_v4l2_buffer(vb);
>>>> +
>>>>            num = scnprintf(str, sizeof(str),
>>>>                    "output [%2d] state = %10s, %8s\n",
>>>>                    i, vb2_stat_name[vb->state],
>>>> @@ -155,11 +162,18 @@ static int vpu_dbg_instance(struct seq_file *s, void *data)
>>>>          vq = v4l2_m2m_get_dst_vq(inst->fh.m2m_ctx);
>>>>        for (i = 0; i < vq->num_buffers; i++) {
>>>> -        struct vb2_buffer *vb = vq->bufs[i];
>>>> -        struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
>>>> +        struct vb2_buffer *vb;
>>>> +        struct vb2_v4l2_buffer *vbuf;
>>>> +
>>>> +        vb = vb2_get_buffer(vq, i);
>>>> +        if (!vb)
>>>> +            continue;
>>>>              if (vb->state == VB2_BUF_STATE_DEQUEUED)
>>>>                continue;
>>>> +
>>>> +        vbuf = to_vb2_v4l2_buffer(vb);
>>>> +
>>>>            num = scnprintf(str, sizeof(str),
>>>>                    "capture[%2d] state = %10s, %8s\n",
>>>>                    i, vb2_stat_name[vb->state],
>>> This can be a separate patch, right? It doesn't depend on any core changes.
>>>
>>> And this can also be applied before this patch.
>> Hans, I would like to clarify this comment (and the following Ditto).
>> Are you against use vb2_get_buffer() outside core ?
>> or testing vb2_get_buffer() result ?
>> The goal of this patch was to remove all access like vq->bufs[i] and to make
>> sure that vb buffer are always valid.
> Sorry for the confusion. I meant that AFAICS each of these driver changes can be
> done in the separate patch and that those separate patches can be applied before
> this patch. I.e., they are independent.
>
> I always prefer specific driver changes to be done as separate patches rather
> than one patch modifying a lot of drivers in one go. That is not always possible,
> of course, but in this case I think it is fine, unless I missed something.

All changes in the drivers use vb2_get_buffer() which is introduced in this patch
so I can't do them before.
If you want I can make a patch per driver but after this patch.

Regards,
Benjamin

>
> Regards,
>
> 	Hans
>
>> Regards,
>> Benjamin
>>
>>>> diff --git a/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c b/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c
>>>> index 621038aab116..62910a1b8a98 100644
>>>> --- a/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c
>>>> +++ b/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c
>>>> @@ -603,7 +603,11 @@ static int mtk_jpeg_qbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
>>>>            return -EINVAL;
>>>>        }
>>>>    -    vb = vq->bufs[buf->index];
>>>> +    vb = vb2_get_buffer(vq, buf->index);
>>>> +    if (!vb) {
>>>> +        dev_err(ctx->jpeg->dev, "buffer not found\n");
>>>> +        return -EINVAL;
>>>> +    }
>>>>        jpeg_src_buf = mtk_jpeg_vb2_to_srcbuf(vb);
>>>>        jpeg_src_buf->bs_size = buf->m.planes[0].bytesused;
>>>>    
>>> Ditto.
>>>
>>>> diff --git a/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_vp9_req_lat_if.c b/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_vp9_req_lat_if.c
>>>> index e393e3e668f8..3d2ae0e1b5b6 100644
>>>> --- a/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_vp9_req_lat_if.c
>>>> +++ b/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_vp9_req_lat_if.c
>>>> @@ -1696,7 +1696,7 @@ static int vdec_vp9_slice_setup_core_buffer(struct vdec_vp9_slice_instance *inst
>>>>          /* update internal buffer's width/height */
>>>>        for (i = 0; i < vq->num_buffers; i++) {
>>>> -        if (vb == vq->bufs[i]) {
>>>> +        if (vb == vb2_get_buffer(vq, i)) {
>>>>                instance->dpb[i].width = w;
>>>>                instance->dpb[i].height = h;
>>>>                break;
>>> Ditto.
>>>
>>>> diff --git a/drivers/media/platform/st/sti/hva/hva-v4l2.c b/drivers/media/platform/st/sti/hva/hva-v4l2.c
>>>> index 3a848ca32a0e..326be09bdb55 100644
>>>> --- a/drivers/media/platform/st/sti/hva/hva-v4l2.c
>>>> +++ b/drivers/media/platform/st/sti/hva/hva-v4l2.c
>>>> @@ -577,6 +577,10 @@ static int hva_qbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
>>>>            }
>>>>              vb2_buf = vb2_get_buffer(vq, buf->index);
>>>> +        if (!vb2_buf) {
>>>> +            dev_dbg(dev, "%s buffer index %d not found\n", ctx->name, buf->index);
>>>> +            return -EINVAL;
>>>> +        }
>>>>            stream = to_hva_stream(to_vb2_v4l2_buffer(vb2_buf));
>>>>            stream->bytesused = buf->bytesused;
>>>>        }
>>> Ditto.
>>>
>>>> diff --git a/drivers/media/test-drivers/visl/visl-dec.c b/drivers/media/test-drivers/visl/visl-dec.c
>>>> index 318d675e5668..ba20ea998d19 100644
>>>> --- a/drivers/media/test-drivers/visl/visl-dec.c
>>>> +++ b/drivers/media/test-drivers/visl/visl-dec.c
>>>> @@ -290,13 +290,20 @@ static void visl_tpg_fill(struct visl_ctx *ctx, struct visl_run *run)
>>>>        for (i = 0; i < out_q->num_buffers; i++) {
>>>>            char entry[] = "index: %u, state: %s, request_fd: %d, ";
>>>>            u32 old_len = len;
>>>> -        char *q_status = visl_get_vb2_state(out_q->bufs[i]->state);
>>>> +        struct vb2_buffer *vb2;
>>>> +        char *q_status;
>>>> +
>>>> +        vb2 = vb2_get_buffer(out_q, i);
>>>> +        if (!vb2)
>>>> +            continue;
>>>> +
>>>> +        q_status = visl_get_vb2_state(vb2->state);
>>>>              len += scnprintf(&buf[len], TPG_STR_BUF_SZ - len,
>>>>                     entry, i, q_status,
>>>> -                 to_vb2_v4l2_buffer(out_q->bufs[i])->request_fd);
>>>> +                 to_vb2_v4l2_buffer(vb2)->request_fd);
>>>>    -        len += visl_fill_bytesused(to_vb2_v4l2_buffer(out_q->bufs[i]),
>>>> +        len += visl_fill_bytesused(to_vb2_v4l2_buffer(vb2),
>>>>                           &buf[len],
>>>>                           TPG_STR_BUF_SZ - len);
>>>>    @@ -342,13 +349,20 @@ static void visl_tpg_fill(struct visl_ctx *ctx, struct visl_run *run)
>>>>        len = 0;
>>>>        for (i = 0; i < cap_q->num_buffers; i++) {
>>>>            u32 old_len = len;
>>>> -        char *q_status = visl_get_vb2_state(cap_q->bufs[i]->state);
>>>> +        struct vb2_buffer *vb2;
>>>> +        char *q_status;
>>>> +
>>>> +        vb2 = vb2_get_buffer(cap_q, i);
>>>> +        if (!vb2)
>>>> +            continue;
>>>> +
>>>> +        q_status = visl_get_vb2_state(vb2->state);
>>>>              len += scnprintf(&buf[len], TPG_STR_BUF_SZ - len,
>>>>                     "index: %u, status: %s, timestamp: %llu, is_held: %d",
>>>> -                 cap_q->bufs[i]->index, q_status,
>>>> -                 cap_q->bufs[i]->timestamp,
>>>> -                 to_vb2_v4l2_buffer(cap_q->bufs[i])->is_held);
>>>> +                 vb2->index, q_status,
>>>> +                 vb2->timestamp,
>>>> +                 to_vb2_v4l2_buffer(vb2)->is_held);
>>>>              tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 16, &buf[old_len]);
>>>>            frame_dprintk(ctx->dev, run->dst->sequence, "%s", &buf[old_len]);
>>> Ditto.
>>>
>>>> diff --git a/drivers/staging/media/atomisp/pci/atomisp_ioctl.c b/drivers/staging/media/atomisp/pci/atomisp_ioctl.c
>>>> index d2174156573a..4b65c69fa60d 100644
>>>> --- a/drivers/staging/media/atomisp/pci/atomisp_ioctl.c
>>>> +++ b/drivers/staging/media/atomisp/pci/atomisp_ioctl.c
>>>> @@ -1061,7 +1061,7 @@ static int atomisp_dqbuf_wrapper(struct file *file, void *fh, struct v4l2_buffer
>>>>        if (ret)
>>>>            return ret;
>>>>    -    vb = pipe->vb_queue.bufs[buf->index];
>>>> +    vb = vb2_get_buffer(&pipe->vb_queue, buf->index);
>>>>        frame = vb_to_frame(vb);
>>>>          buf->reserved = asd->frame_status[buf->index];
>>> Ditto.
>>>
>>> Background: I think it is really useful to merge a lot of the groundwork early
>>> on, where possible. It simplifies the remainder of the patch series.
>>>
>>> Regards,
>>>
>>>      Hans
>>>