Re: [PATCH] powerpc/mpic: Fix memory leak of mpic on error paths

[Jyoti Gupta]

On 2026-09-29 13:02, Amit Machhiwal wrote:
Hi Jyoti,
For some reason, you response to this thread is not appearing on the 
list.  I
think you mail client is messing up the "In-Reply-To" information in 
the
header.  Fix look into fixing it.

For now, I'll qoute your reply here with my further response.

Hello Amit ,

I'm not sure why my reply did not reach the list,but it is visible at 
the link below:
https://lore.kernel.org/all/e3c17fc50f7897b08fbd67a120087...@linux.ibm.com/

let me know if this looks fine.

On 2026/09/24 03:05 PM, Amit Machhiwal wrote:
On 2026/09/24 12:00 PM, Jyoti Gupta wrote:
> mpic_alloc() allocates struct mpic (and optionally the protected
> sources bitmap) but does not free them when mpic_setup_error_int() or
> irq_domain_create_linear() fails, leaking the memory.
>
> Initialise mpic to NULL and free the bitmap and the mpic in the common
> error path before dropping the device node reference. The register
> mappings are not released here.
>
> Signed-off-by: Jyoti Gupta <jy...@linux.ibm.com>

I see a couple of problems with this patch:

1. The patch does not apply.  Both "git apply" and "git am" fail on 
this patch.
   I think the context lines around the irq_domain_create_linear() 
call have
   tabs converted to spaces, which causes the context match to fail.

      ❯ git apply 
./20260924_jyoti_powerpc_mpic_fix_memory_leak_of_mpic_on_error_paths.mbx 
--check --verbose
      Checking patch arch/powerpc/sysdev/mpic.c...
      error: while searching for:
                                                 intvec_top,
                                                 &mpic_host_ops, mpic);

        /*
      	 * FIXME: The code leaks the MPIC object and mappings here; 
this
         * is very unlikely to fail but it ought to be fixed anyways.
         */
        if (mpic->irqhost == NULL)
                goto err_of_node_put;


      error: patch failed: arch/powerpc/sysdev/mpic.c:1488
      error: arch/powerpc/sysdev/mpic.c: patch does not apply

   Please regenerate with:

     git format-patch -1

   and verify with:

     git apply --check <patch>

   before resending.

2. The mpic_setup_error_int() leak is not actually fixed.  The commit 
message
   says both mpic_setup_error_int() and irq_domain_create_linear() 
failure paths
   are fixed, but the former still does a bare return NULL:

     if (fsl_version >= 0x401) {
         ret = mpic_setup_error_int(mpic, intvec_top - 13);
         if (ret)
             return NULL;   /* <-- still leaks mpic and 
mpic->protected */
     }

   This needs to become:

         if (ret)
             goto err_of_node_put;

Thanks,
Amit

> ---
>  arch/powerpc/sysdev/mpic.c | 9 ++++-----
>  1 file changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/arch/powerpc/sysdev/mpic.c b/arch/powerpc/sysdev/mpic.c
> index 6e8fa94985f3..e777b33f170d 100644
> --- a/arch/powerpc/sysdev/mpic.c
> +++ b/arch/powerpc/sysdev/mpic.c
> @@ -1221,7 +1221,7 @@ struct mpic * __init mpic_alloc(struct device_node 
*node,
>                            const char *name)
>  {
>    int i, psize, intvec_top;
> -  struct mpic *mpic;
> +  struct mpic *mpic = NULL;
>    u32 greg_feature;
>    const char *vers;
>    const u32 *psrc;
> @@ -1488,10 +1488,6 @@ struct mpic * __init mpic_alloc(struct device_node 
*node,
>                                             intvec_top,
>                                             &mpic_host_ops, mpic);
>
> -  /*
> -   * FIXME: The code leaks the MPIC object and mappings here; this
> -   * is very unlikely to fail but it ought to be fixed anyways.
> -   */
>    if (mpic->irqhost == NULL)
>            goto err_of_node_put;
>
> @@ -1527,6 +1523,9 @@ struct mpic * __init mpic_alloc(struct device_node 
*node,
>    return mpic;
>
>  err_of_node_put:
> +  if (mpic)
> +          bitmap_free(mpic->protected);
> +  kfree(mpic);
>    of_node_put(node);
>    return NULL;
>  }
> --
> 2.55.0
>
>


Response from Jyoti:

The mpic_setup_error_int() leak is intentionally left to a separate 
patch
already under

If that's the case, you shouldn't have mentioned about this change in 
the commit
log of this patch.  The commit log needs to be updated accordingly if 
that's
what your intent is.

Agree,I'llfix it in v2.

 review:
"[PATCH] powerpc: Fix missing of_node_put() calls"
That patch converts the return NULL at line 1392 to goto 
err_of_node_put as
part of a

I could only find below:

https://lore.kernel.org/all/5cef10ab1610113fe162cec1dffd9...@linux.ibm.com/

Sorry, but I don't think thats the right way to send a patch.  You seem 
to have
replied to the email which did not land into the mailing lists at all.

Yes, I made a mistake earlier by not keeping 
linuxppc-dev@lists.ozlabs.org
in the loop. I later replied to all with the list included, but I 
understand
that was not the right way to do it. I'll resend the patch as a fresh 
thread
in the proper format.

Having
said that, I think fixing this particular path in mpic_alloc() would 
make more
sense in the current patch itself.

sure we can do that too.

Please look into fixing your mail client and send the patches/responses 
in
proper format.


Got it.

Thanks,
Jyoti

Thanks,
Amit.

broader of_node_put() fix series. This patch focuses only on the
irq_domain_create_linear()
failure path.

Thanks,
Jyoti