In cases where mapping of mpmu/apmu/apbc registers fails, the code path
does not handle the failure gracefully, potentially leading to a memory
leak. This fix ensures proper cleanup by freeing the allocated memory
for 'pxa_unit' before returning.
Signed-off-by: Kuan-Wei Chiu <[email protected]>
---
drivers/clk/mmp/clk-of-pxa168.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/clk/mmp/clk-of-pxa168.c b/drivers/clk/mmp/clk-of-pxa168.c
index fb0df64cf053..c5a7ba1deaa3 100644
--- a/drivers/clk/mmp/clk-of-pxa168.c
+++ b/drivers/clk/mmp/clk-of-pxa168.c
@@ -308,18 +308,21 @@ static void __init pxa168_clk_init(struct device_node *np)
pxa_unit->mpmu_base = of_iomap(np, 0);
if (!pxa_unit->mpmu_base) {
pr_err("failed to map mpmu registers\n");
+ kfree(pxa_unit);
return;
}
pxa_unit->apmu_base = of_iomap(np, 1);
if (!pxa_unit->apmu_base) {
pr_err("failed to map apmu registers\n");
+ kfree(pxa_unit);
return;
}
pxa_unit->apbc_base = of_iomap(np, 2);
if (!pxa_unit->apbc_base) {
pr_err("failed to map apbc registers\n");
+ kfree(pxa_unit);
return;
}
--
2.25.1
On Mon, Dec 11, 2023 at 01:52:32AM +0800, Kuan-Wei Chiu wrote:
> In cases where mapping of mpmu/apmu/apbc registers fails, the code path
> does not handle the failure gracefully, potentially leading to a memory
> leak. This fix ensures proper cleanup by freeing the allocated memory
> for 'pxa_unit' before returning.
>
> Signed-off-by: Kuan-Wei Chiu <[email protected]>
I dunno why you have CCed me on this tbh, I guess I must have acked a
patch related to this driver's bindings somehow.
I wonder, is there actually any point in freeing these, isn't the system
going to fail to actually boot if these allocations fail?
> ---
> drivers/clk/mmp/clk-of-pxa168.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/clk/mmp/clk-of-pxa168.c b/drivers/clk/mmp/clk-of-pxa168.c
> index fb0df64cf053..c5a7ba1deaa3 100644
> --- a/drivers/clk/mmp/clk-of-pxa168.c
> +++ b/drivers/clk/mmp/clk-of-pxa168.c
> @@ -308,18 +308,21 @@ static void __init pxa168_clk_init(struct device_node *np)
> pxa_unit->mpmu_base = of_iomap(np, 0);
> if (!pxa_unit->mpmu_base) {
> pr_err("failed to map mpmu registers\n");
> + kfree(pxa_unit);
> return;
> }
>
> pxa_unit->apmu_base = of_iomap(np, 1);
> if (!pxa_unit->apmu_base) {
> pr_err("failed to map apmu registers\n");
> + kfree(pxa_unit);
> return;
> }
>
> pxa_unit->apbc_base = of_iomap(np, 2);
> if (!pxa_unit->apbc_base) {
> pr_err("failed to map apbc registers\n");
> + kfree(pxa_unit);
> return;
> }
>
> --
> 2.25.1
>
On Mon, Dec 11, 2023 at 01:19:17PM +0000, Conor Dooley wrote:
> On Mon, Dec 11, 2023 at 01:52:32AM +0800, Kuan-Wei Chiu wrote:
> > In cases where mapping of mpmu/apmu/apbc registers fails, the code path
> > does not handle the failure gracefully, potentially leading to a memory
> > leak. This fix ensures proper cleanup by freeing the allocated memory
> > for 'pxa_unit' before returning.
> >
> > Signed-off-by: Kuan-Wei Chiu <[email protected]>
>
> I dunno why you have CCed me on this tbh, I guess I must have acked a
> patch related to this driver's bindings somehow.
>
I included you in the CC list based on the results from
./scripts/get_maintainer.pl, which listed your email. I apologize for
any inconvenience. I will be more careful in selecting the recipients
for my patches in the future.
> I wonder, is there actually any point in freeing these, isn't the system
> going to fail to actually boot if these allocations fail?
>
I'm not familiar with the boot process. I noticed similar cleanup in
related code (e.g., clk-of-pxa1928.c) and thought it might be an
oversight in pxa168. If releasing memory here isn't essential, I'm
inclined to drop the patch.
However, if the cleanup is necessary, I believe additional iounmap,
similar to commit 19b9f29 (clk:mmp:clk-of-pxa910: Free memory and
Unmap region obtained by kzmalloc and of_iomap), may be required.
If necessary, I'm willing to send a v2 patch with the additional
changes.
Best regards,
Kuan-Wei Chiu
> > ---
> > drivers/clk/mmp/clk-of-pxa168.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/clk/mmp/clk-of-pxa168.c b/drivers/clk/mmp/clk-of-pxa168.c
> > index fb0df64cf053..c5a7ba1deaa3 100644
> > --- a/drivers/clk/mmp/clk-of-pxa168.c
> > +++ b/drivers/clk/mmp/clk-of-pxa168.c
> > @@ -308,18 +308,21 @@ static void __init pxa168_clk_init(struct device_node *np)
> > pxa_unit->mpmu_base = of_iomap(np, 0);
> > if (!pxa_unit->mpmu_base) {
> > pr_err("failed to map mpmu registers\n");
> > + kfree(pxa_unit);
> > return;
> > }
> >
> > pxa_unit->apmu_base = of_iomap(np, 1);
> > if (!pxa_unit->apmu_base) {
> > pr_err("failed to map apmu registers\n");
> > + kfree(pxa_unit);
> > return;
> > }
> >
> > pxa_unit->apbc_base = of_iomap(np, 2);
> > if (!pxa_unit->apbc_base) {
> > pr_err("failed to map apbc registers\n");
> > + kfree(pxa_unit);
> > return;
> > }
> >
> > --
> > 2.25.1
> >
Quoting Kuan-Wei Chiu (2023-12-10 09:52:32)
> In cases where mapping of mpmu/apmu/apbc registers fails, the code path
> does not handle the failure gracefully, potentially leading to a memory
> leak. This fix ensures proper cleanup by freeing the allocated memory
> for 'pxa_unit' before returning.
>
> Signed-off-by: Kuan-Wei Chiu <[email protected]>
> ---
Applied to clk-next