Hello Yong Zhi,
The patch 7fc7af649ca7: "media: staging/intel-ipu3: Add imgu top
level pci device driver" from Dec 6, 2018, leads to the following
static checker warning:
drivers/staging/media/ipu3/ipu3.c:493 imgu_isr_threaded()
warn: 'b' is an error pointer or valid
drivers/staging/media/ipu3/ipu3.c
472 static irqreturn_t imgu_isr_threaded(int irq, void *imgu_ptr)
473 {
474 struct imgu_device *imgu = imgu_ptr;
475 struct imgu_media_pipe *imgu_pipe;
476 int p;
477
478 /* Dequeue / queue buffers */
479 do {
480 u64 ns = ktime_get_ns();
481 struct ipu3_css_buffer *b;
482 struct imgu_buffer *buf = NULL;
483 unsigned int node, pipe;
484 bool dummy;
485
486 do {
487 mutex_lock(&imgu->lock);
488 b = ipu3_css_buf_dequeue(&imgu->css);
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ipu3_css_buf_dequeue() doesn't return NULL.
489 mutex_unlock(&imgu->lock);
490 } while (PTR_ERR(b) == -EAGAIN);
491
492 if (IS_ERR_OR_NULL(b)) {
^^^^^^^^^^^^^^^^^
--> 493 if (!b || PTR_ERR(b) == -EBUSY) /* All done */
^^
When a function returns both NULL and error pointers, then NULL is
considered a special case of success. Like perhaps you request a
feature, but that feature isn't enabled in the config. It's fine,
because the user *chose* to turn off the feature, so it's not an error
but we also don't have a valid pointer we can use.
It looks like you were probably trying to do something like that but
you missed part of the commit? Otherwise we should delete the dead
code.
494 break;
495 dev_err(&imgu->pci_dev->dev,
496 "failed to dequeue buffers (%ld)\n",
497 PTR_ERR(b));
498 break;
499 }
500
regards,
dan carpenter
