Re: [PATCH] fbdev: platinumfb: add error checking for ioremap calls

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] The return value of `request_mem_region` for `cmap_regs_phys` is ignored, and the error path unconditionally calls `release_mem_region` on it.
--

--- Patch [1]: [PATCH] fbdev: platinumfb: add error checking for ioremap calls ---
commit 71a977c44688a9c49d5d28cb513b56fa5a810a1d
Author: yuebingkun <[email protected]>

fbdev: platinumfb: add error checking for ioremap calls

This commit adds error checking for ioremap() and ioremap_wt() calls in
platinumfb_probe() to prevent NULL pointer dereferences. It also transitions
the function to use goto-based cleanup for handling initialization failures.

> diff --git a/drivers/video/fbdev/platinumfb.c b/drivers/video/fbdev/platinumfb.c
> index a08d955d9b439..f74c937c839f8 100644
> --- a/drivers/video/fbdev/platinumfb.c
> +++ b/drivers/video/fbdev/platinumfb.c

[ ... ]

> @@ -623,13 +638,22 @@ static int platinumfb_probe(struct platform_device* odev)
>  	dev_set_drvdata(&odev->dev, info);
>  
>  	rc = platinum_init_fb(info);
> -	if (rc != 0) {
> -		iounmap(pinfo->frame_buffer);
> -		iounmap(pinfo->platinum_regs);
> -		iounmap(pinfo->cmap_regs);
> -		framebuffer_release(info);
> -	}
> +	if (rc != 0)
> +		goto err_unmap_cmap;
> +
> +	return 0;
>  
> +err_unmap_cmap:
> +	iounmap(pinfo->cmap_regs);
> +err_release_cmap:
> +	release_mem_region(pinfo->cmap_regs_phys, 0x1000);

[Severity: High]
This is a pre-existing issue, but by adding this new error path, the code
propagates an unsafe assumption.

Earlier in platinumfb_probe(), the return value of request_mem_region()
is ignored:

	request_mem_region(pinfo->cmap_regs_phys, 0x1000, "platinumfb cmap");
	pinfo->cmap_regs = ioremap(pinfo->cmap_regs_phys, 0x1000);

If request_mem_region() fails (for instance, if the region is already
claimed), execution continues. If a subsequent step like ioremap() or
platinum_init_fb() then fails, the code will jump to this newly added
err_release_cmap label and unconditionally release the region.

Because __release_region() searches by address and doesn't verify the caller,
could this forcefully free an I/O memory region that actually belongs to
another driver, potentially leading to resource tree corruption or a
use-after-free?

> +err_unmap_regs:
> +	iounmap(pinfo->platinum_regs);
> +err_unmap_fb:
> +	iounmap(pinfo->frame_buffer);

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1