diff mbox

dmaengine: usb-dmac: fix endless loop in usb_dmac_chan_terminate_all()

Message ID 1447303060-7880-1-git-send-email-yoshihiro.shimoda.uh@renesas.com (mailing list archive)
State Accepted
Delegated to: Geert Uytterhoeven
Headers show

Commit Message

Yoshihiro Shimoda Nov. 12, 2015, 4:37 a.m. UTC 
This patch fixes an issue that list_for_each_entry() in
usb_dmac_chan_terminate_all() is possible to cause endless loop because
this will move own desc to the desc_freed. So, this driver should use
list_for_each_entry_safe() instead of list_for_each_entry().

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
---
 drivers/dma/sh/usb-dmac.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Vinod Koul Nov. 16, 2015, 3:46 a.m. UTC | #1 
On Thu, Nov 12, 2015 at 01:37:40PM +0900, Yoshihiro Shimoda wrote:
> This patch fixes an issue that list_for_each_entry() in
> usb_dmac_chan_terminate_all() is possible to cause endless loop because
> this will move own desc to the desc_freed. So, this driver should use
> list_for_each_entry_safe() instead of list_for_each_entry().

Sorry this is not clear to my why this owuld have endless loop, can you
explain more

> 
> Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
> ---
>  drivers/dma/sh/usb-dmac.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/dma/sh/usb-dmac.c b/drivers/dma/sh/usb-dmac.c
> index ebd8a5f..16fb330 100644
> --- a/drivers/dma/sh/usb-dmac.c
> +++ b/drivers/dma/sh/usb-dmac.c
> @@ -448,7 +448,7 @@ usb_dmac_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
>  static int usb_dmac_chan_terminate_all(struct dma_chan *chan)
>  {
>  	struct usb_dmac_chan *uchan = to_usb_dmac_chan(chan);
> -	struct usb_dmac_desc *desc;
> +	struct usb_dmac_desc *desc, *_desc;
>  	unsigned long flags;
>  	LIST_HEAD(head);
>  	LIST_HEAD(list);
> @@ -459,7 +459,7 @@ static int usb_dmac_chan_terminate_all(struct dma_chan *chan)
>  	if (uchan->desc)
>  		uchan->desc = NULL;
>  	list_splice_init(&uchan->desc_got, &list);
> -	list_for_each_entry(desc, &list, node)
> +	list_for_each_entry_safe(desc, _desc, &list, node)
>  		list_move_tail(&desc->node, &uchan->desc_freed);
>  	spin_unlock_irqrestore(&uchan->vc.lock, flags);
>  	vchan_dma_desc_free_list(&uchan->vc, &head);
> -- 
> 1.9.1
> 
> --
> To unsubscribe from this list: send the line "unsubscribe dmaengine" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
Yoshihiro Shimoda Nov. 16, 2015, 7:13 a.m. UTC | #2 
Hi Vinod,

> From: Vinod Koul [mailto:vinod.koul@intel.com]
> Sent: Monday, November 16, 2015 12:47 PM
> 
> On Thu, Nov 12, 2015 at 01:37:40PM +0900, Yoshihiro Shimoda wrote:
> > This patch fixes an issue that list_for_each_entry() in
> > usb_dmac_chan_terminate_all() is possible to cause endless loop because
> > this will move own desc to the desc_freed. So, this driver should use
> > list_for_each_entry_safe() instead of list_for_each_entry().
> 
> Sorry this is not clear to my why this owuld have endless loop, can you
> explain more

This issue will happens if &uchan->desc_got has one or more lists.
(In other words, there was a potential problem from the first commit of this driver...)

This is because that this driver use "desc" for list_move_tail() as the first
argument, and the list_move_tail() calls __list_del_entry(). So, the desc may be
wrong list_head for list_next_entry().

static int usb_dmac_chan_terminate_all(struct dma_chan *chan)
{
<snip>
  	list_splice_init(&uchan->desc_got, &list);
	list_for_each_entry(desc, &list, node)
  		list_move_tail(&desc->node, &uchan->desc_freed);

< in include/linux/list.h >
static inline void list_move_tail(struct list_head *list,
                                  struct list_head *head)
{
        __list_del_entry(list);
        list_add_tail(list, head);
}

If I enabled CONFIG_DEBUG_LIST, the following WARNING happened:
------------[ cut here ]------------
WARNING: CPU: 2 PID: 2477 at lib/list_debug.c:33 __list_add+0x64/0xb0()
list_add corruption. prev->next should be next (edca6098), but was ee936e00. (prev=ee936ec0).
Modules linked in: g_mass_storage(-) usb_f_mass_storage libcomposite
CPU: 2 PID: 2477 Comm: file-storage Not tainted 4.3.0-rc1-00002-g4d15d22 #27
Hardware name: Generic R8A7790 (Flattened Device Tree)
Backtrace:
[<c00137a0>] (dump_backtrace) from [<c0013998>] (show_stack+0x18/0x1c)
 r7:c05d2664 r6:ee92e140 r5:00000009 r4:00000000
[<c0013980>] (show_stack) from [<c01be7d0>] (dump_stack+0x78/0x94)
[<c01be758>] (dump_stack) from [<c0025224>] (warn_slowpath_common+0x8c/0xb8)
 r5:00000009 r4:ee22bd90
[<c0025198>] (warn_slowpath_common) from [<c0025288>] (warn_slowpath_fmt+0x38/0x40)
 r9:edca6098 r8:a0000093 r7:edca605c r6:edca6098 r5:edca6098 r4:ee936ec0
[<c0025254>] (warn_slowpath_fmt) from [<c01d49b8>] (__list_add+0x64/0xb0)
 r3:edca6098 r2:c05d2675
[<c01d4954>] (__list_add) from [<c0202e5c>] (usb_dmac_chan_terminate_all+0x14c/0x17c)
 r6:ee22bde0 r5:edca6058 r4:edca6010
[<c0202d10>] (usb_dmac_chan_terminate_all) from [<c02e84b0>] (usbhs_pkt_pop+0xa0/0xf4)
< snip >

Best regards,
Yoshihiro Shimoda

> >
> > Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
> > ---
> >  drivers/dma/sh/usb-dmac.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/dma/sh/usb-dmac.c b/drivers/dma/sh/usb-dmac.c
> > index ebd8a5f..16fb330 100644
> > --- a/drivers/dma/sh/usb-dmac.c
> > +++ b/drivers/dma/sh/usb-dmac.c
> > @@ -448,7 +448,7 @@ usb_dmac_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
> >  static int usb_dmac_chan_terminate_all(struct dma_chan *chan)
> >  {
> >  	struct usb_dmac_chan *uchan = to_usb_dmac_chan(chan);
> > -	struct usb_dmac_desc *desc;
> > +	struct usb_dmac_desc *desc, *_desc;
> >  	unsigned long flags;
> >  	LIST_HEAD(head);
> >  	LIST_HEAD(list);
> > @@ -459,7 +459,7 @@ static int usb_dmac_chan_terminate_all(struct dma_chan *chan)
> >  	if (uchan->desc)
> >  		uchan->desc = NULL;
> >  	list_splice_init(&uchan->desc_got, &list);
> > -	list_for_each_entry(desc, &list, node)
> > +	list_for_each_entry_safe(desc, _desc, &list, node)
> >  		list_move_tail(&desc->node, &uchan->desc_freed);
> >  	spin_unlock_irqrestore(&uchan->vc.lock, flags);
> >  	vchan_dma_desc_free_list(&uchan->vc, &head);
> > --
> > 1.9.1
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe dmaengine" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> --
> ~Vinod
--
To unsubscribe from this list: send the line "unsubscribe linux-sh" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Vinod Koul Dec. 5, 2015, 8:08 a.m. UTC | #3 
On Thu, Nov 12, 2015 at 01:37:40PM +0900, Yoshihiro Shimoda wrote:
> This patch fixes an issue that list_for_each_entry() in
> usb_dmac_chan_terminate_all() is possible to cause endless loop because
> this will move own desc to the desc_freed. So, this driver should use
> list_for_each_entry_safe() instead of list_for_each_entry().

Applied now, thanks
diff mbox

Patch

diff --git a/drivers/dma/sh/usb-dmac.c b/drivers/dma/sh/usb-dmac.c
index ebd8a5f..16fb330 100644
--- a/drivers/dma/sh/usb-dmac.c
+++ b/drivers/dma/sh/usb-dmac.c
@@ -448,7 +448,7 @@  usb_dmac_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 static int usb_dmac_chan_terminate_all(struct dma_chan *chan)
 {
 	struct usb_dmac_chan *uchan = to_usb_dmac_chan(chan);
-	struct usb_dmac_desc *desc;
+	struct usb_dmac_desc *desc, *_desc;
 	unsigned long flags;
 	LIST_HEAD(head);
 	LIST_HEAD(list);
@@ -459,7 +459,7 @@  static int usb_dmac_chan_terminate_all(struct dma_chan *chan)
 	if (uchan->desc)
 		uchan->desc = NULL;
 	list_splice_init(&uchan->desc_got, &list);
-	list_for_each_entry(desc, &list, node)
+	list_for_each_entry_safe(desc, _desc, &list, node)
 		list_move_tail(&desc->node, &uchan->desc_freed);
 	spin_unlock_irqrestore(&uchan->vc.lock, flags);
 	vchan_dma_desc_free_list(&uchan->vc, &head);