2014-01-26 11:01:09

by Jose Alonso

[permalink] [raw]
Subject: [PATCH] for_each macros correctness


I observed that there are for_each macros that do an extra memory access
beyond the defined area.
Normally this does not cause problems.
But, this can cause exceptions. For example: if the area is allocated at
the end of a page and the next page is not accessible.

For correctness, I suggest changing the arguments of the 'for loop' like
others 'for_each' do in the kernel.

files involved:
drivers/s390/cio/qdio.h
drivers/scsi/isci/host.h
drivers/sh/clk/core.c
include/linux/blk-mq.h
include/linux/shdma-base.h
sound/soc/sh/rcar/adg.c


Signed-off-by: Jose Alonso <[email protected]>
---
diff --git a/drivers/s390/cio/qdio.h b/drivers/s390/cio/qdio.h
index 8acaae1..a563e4c 100644
--- a/drivers/s390/cio/qdio.h
+++ b/drivers/s390/cio/qdio.h
@@ -359,14 +359,12 @@ static inline int multicast_outbound(struct qdio_q *q)
#define need_siga_sync_out_after_pci(q) \
(unlikely(q->irq_ptr->siga_flag.sync_out_after_pci))

-#define for_each_input_queue(irq_ptr, q, i) \
- for (i = 0, q = irq_ptr->input_qs[0]; \
- i < irq_ptr->nr_input_qs; \
- q = irq_ptr->input_qs[++i])
-#define for_each_output_queue(irq_ptr, q, i) \
- for (i = 0, q = irq_ptr->output_qs[0]; \
- i < irq_ptr->nr_output_qs; \
- q = irq_ptr->output_qs[++i])
+#define for_each_input_queue(irq_ptr, q, i) \
+ for (i = 0; i < irq_ptr->nr_input_qs && \
+ ({ q = irq_ptr->input_qs[i]; 1; }); i++)
+#define for_each_output_queue(irq_ptr, q, i) \
+ for (i = 0; i < irq_ptr->nr_output_qs && \
+ ({ q = irq_ptr->output_qs[i]; 1; }); i++)

#define prev_buf(bufnr) \
((bufnr + QDIO_MAX_BUFFERS_MASK) & QDIO_MAX_BUFFERS_MASK)
diff --git a/drivers/scsi/isci/host.h b/drivers/scsi/isci/host.h
index 4911310..99c5a69 100644
--- a/drivers/scsi/isci/host.h
+++ b/drivers/scsi/isci/host.h
@@ -311,9 +311,8 @@ static inline struct Scsi_Host *to_shost(struct isci_host *ihost)
}

#define for_each_isci_host(id, ihost, pdev) \
- for (id = 0, ihost = to_pci_info(pdev)->hosts[id]; \
- id < ARRAY_SIZE(to_pci_info(pdev)->hosts) && ihost; \
- ihost = to_pci_info(pdev)->hosts[++id])
+ for (id = 0; id < ARRAY_SIZE(to_pci_info(pdev)->hosts) && \
+ (ihost = to_pci_info(pdev)->hosts[id]); id++)

static inline void wait_for_start(struct isci_host *ihost)
{
diff --git a/drivers/sh/clk/core.c b/drivers/sh/clk/core.c
index 7472785..6dd3cbb 100644
--- a/drivers/sh/clk/core.c
+++ b/drivers/sh/clk/core.c
@@ -82,8 +82,8 @@ struct clk_rate_round_data {
};

#define for_each_frequency(pos, r, freq) \
- for (pos = r->min, freq = r->func(pos, r); \
- pos <= r->max; pos++, freq = r->func(pos, r)) \
+ for (pos = r->min; pos <= r->max && \
+ ({ freq = r->func(pos, r); 1; }); pos++) \
if (unlikely(freq == 0)) \
; \
else
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index ab0e9b2..a4b40cf 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -159,16 +159,16 @@ static inline struct request *blk_mq_tag_to_rq(struct blk_mq_hw_ctx *hctx,
}

#define queue_for_each_hw_ctx(q, hctx, i) \
- for ((i) = 0, hctx = (q)->queue_hw_ctx[0]; \
- (i) < (q)->nr_hw_queues; (i)++, hctx = (q)->queue_hw_ctx[i])
+ for ((i) = 0; (i) < (q)->nr_hw_queues && \
+ ({ hctx = (q)->queue_hw_ctx[i]; 1; }); (i)++)

#define queue_for_each_ctx(q, ctx, i) \
- for ((i) = 0, ctx = per_cpu_ptr((q)->queue_ctx, 0); \
- (i) < (q)->nr_queues; (i)++, ctx = per_cpu_ptr(q->queue_ctx, (i)))
+ for ((i) = 0; (i) < (q)->nr_queues && \
+ ({ ctx = per_cpu_ptr((q)->queue_ctx, (i)); 1; }); (i)++)

#define hctx_for_each_ctx(hctx, ctx, i) \
- for ((i) = 0, ctx = (hctx)->ctxs[0]; \
- (i) < (hctx)->nr_ctx; (i)++, ctx = (hctx)->ctxs[(i)])
+ for ((i) = 0; (i) < (hctx)->nr_ctx && \
+ ({ ctx = (hctx)->ctxs[(i)]; 1; }); (i)++)

#define blk_ctx_sum(q, sum) \
({ \
diff --git a/include/linux/shdma-base.h b/include/linux/shdma-base.h
index f92c0a4..35285f7 100644
--- a/include/linux/shdma-base.h
+++ b/include/linux/shdma-base.h
@@ -111,8 +111,8 @@ struct shdma_dev {
size_t desc_size;
};

-#define shdma_for_each_chan(c, d, i) for (i = 0, c = (d)->schan[0]; \
- i < (d)->dma_dev.chancnt; c = (d)->schan[++i])
+#define shdma_for_each_chan(c, d, i) for (i = 0; i < (d)->dma_dev.chancnt && \
+ ({ c = (d)->schan[i]; 1; }); i++)

int shdma_request_irq(struct shdma_chan *, int,
unsigned long, const char *);
diff --git a/sound/soc/sh/rcar/adg.c b/sound/soc/sh/rcar/adg.c
index a53235c..ef19c15 100644
--- a/sound/soc/sh/rcar/adg.c
+++ b/sound/soc/sh/rcar/adg.c
@@ -25,9 +25,8 @@ struct rsnd_adg {
};

#define for_each_rsnd_clk(pos, adg, i) \
- for (i = 0, (pos) = adg->clk[i]; \
- i < CLKMAX; \
- i++, (pos) = adg->clk[i])
+ for (i = 0; i < CLKMAX && \
+ ({ (pos) = adg->clk[i]; 1; }); i++)
#define rsnd_priv_to_adg(priv) ((struct rsnd_adg *)(priv)->adg)

static int rsnd_adg_set_convert_clk_gen1(struct rsnd_priv *priv,


2014-01-26 13:39:40

by Fubo Chen

[permalink] [raw]
Subject: Re: [PATCH] for_each macros correctness

On Sun, Jan 26, 2014 at 11:54 AM, Jose Alonso <[email protected]> wrote:
> I observed that there are for_each macros that do an extra memory access
> beyond the defined area.
> Normally this does not cause problems.
> But, this can cause exceptions. For example: if the area is allocated at
> the end of a page and the next page is not accessible.
>
> For correctness, I suggest changing the arguments of the 'for loop' like
> others 'for_each' do in the kernel.

Does this patch fix a kernel crash when using gcc 4.8 like the patch
in http://lkml.org/lkml/2014/1/21/146 ?

Thanks,

Fubo.

2014-01-26 18:05:14

by Jose Alonso

[permalink] [raw]
Subject: Re: [PATCH] for_each macros correctness

On Sun, 2014-01-26 at 14:39 +0100, Fubo Chen wrote:
> On Sun, Jan 26, 2014 at 11:54 AM, Jose Alonso <[email protected]> wrote:
> > I observed that there are for_each macros that do an extra memory access
> > beyond the defined area.
> > Normally this does not cause problems.
> > But, this can cause exceptions. For example: if the area is allocated at
> > the end of a page and the next page is not accessible.
> >
> > For correctness, I suggest changing the arguments of the 'for loop' like
> > others 'for_each' do in the kernel.
>
> Does this patch fix a kernel crash when using gcc 4.8 like the patch
> in http://lkml.org/lkml/2014/1/21/146 ?

The patch for 'for_each_isci_host' is equivalent.

I followed the link above, and as I understand: The access to
to_pci_info(pdev)->hosts[2] do not caused exception, but
the fact of hosts[2] is "logically accessed" "confused" the
gcc 4.8 compiler. (The gcc compiler do very aggressive loop
optimization and can eliminate loops executing each pass
linearly).

2014-01-27 09:26:53

by Dorau, Lukasz

[permalink] [raw]
Subject: RE: [PATCH] for_each macros correctness

On Sunday, January 26, 2014 11:54 AM Jose Alonso <[email protected]> wrote:
> I observed that there are for_each macros that do an extra memory access
> beyond the defined area.
> Normally this does not cause problems.
> But, this can cause exceptions. For example: if the area is allocated at
> the end of a page and the next page is not accessible.
>
> For correctness, I suggest changing the arguments of the 'for loop' like
> others 'for_each' do in the kernel.
>
> files involved:
> drivers/s390/cio/qdio.h
> drivers/scsi/isci/host.h

Hi Jose,

The macro in "drivers/scsi/isci/host.h" has already been corrected:
http://marc.info/?l=linux-kernel&m=139030404415970&w=2
So you can remove the part regarding "drivers/scsi/isci/host.h" from your patch.

Lukasz

????{.n?+???????+%?????ݶ??w??{.n?+????{??G?????{ay?ʇڙ?,j??f???h?????????z_??(?階?ݢj"???m??????G????????????&???~???iO???z??v?^?m???? ????????I?

2014-01-27 20:02:13

by Jens Axboe

[permalink] [raw]
Subject: Re: [PATCH] for_each macros correctness

On Sun, Jan 26 2014, Jose Alonso wrote:
>
> I observed that there are for_each macros that do an extra memory access
> beyond the defined area.
> Normally this does not cause problems.
> But, this can cause exceptions. For example: if the area is allocated at
> the end of a page and the next page is not accessible.
>
> For correctness, I suggest changing the arguments of the 'for loop' like
> others 'for_each' do in the kernel.
>
> files involved:
> drivers/s390/cio/qdio.h
> drivers/scsi/isci/host.h
> drivers/sh/clk/core.c
> include/linux/blk-mq.h
> include/linux/shdma-base.h
> sound/soc/sh/rcar/adg.c

Thanks, I'll dig out the blk-mq bit.

--
Jens Axboe

2014-01-28 12:02:53

by Heiko Carstens

[permalink] [raw]
Subject: Re: [PATCH] for_each macros correctness

On Sun, Jan 26, 2014 at 08:54:18AM -0200, Jose Alonso wrote:
>
> I observed that there are for_each macros that do an extra memory access
> beyond the defined area.
> Normally this does not cause problems.
> But, this can cause exceptions. For example: if the area is allocated at
> the end of a page and the next page is not accessible.
>
> For correctness, I suggest changing the arguments of the 'for loop' like
> others 'for_each' do in the kernel.
>
> files involved:
> drivers/s390/cio/qdio.h
> drivers/scsi/isci/host.h
> drivers/sh/clk/core.c
> include/linux/blk-mq.h
> include/linux/shdma-base.h
> sound/soc/sh/rcar/adg.c
>
>
> Signed-off-by: Jose Alonso <[email protected]>
> ---
> diff --git a/drivers/s390/cio/qdio.h b/drivers/s390/cio/qdio.h
> index 8acaae1..a563e4c 100644
> --- a/drivers/s390/cio/qdio.h
> +++ b/drivers/s390/cio/qdio.h
> @@ -359,14 +359,12 @@ static inline int multicast_outbound(struct qdio_q *q)
> #define need_siga_sync_out_after_pci(q) \
> (unlikely(q->irq_ptr->siga_flag.sync_out_after_pci))
>
> -#define for_each_input_queue(irq_ptr, q, i) \
> - for (i = 0, q = irq_ptr->input_qs[0]; \
> - i < irq_ptr->nr_input_qs; \
> - q = irq_ptr->input_qs[++i])
> -#define for_each_output_queue(irq_ptr, q, i) \
> - for (i = 0, q = irq_ptr->output_qs[0]; \
> - i < irq_ptr->nr_output_qs; \
> - q = irq_ptr->output_qs[++i])
> +#define for_each_input_queue(irq_ptr, q, i) \
> + for (i = 0; i < irq_ptr->nr_input_qs && \
> + ({ q = irq_ptr->input_qs[i]; 1; }); i++)
> +#define for_each_output_queue(irq_ptr, q, i) \
> + for (i = 0; i < irq_ptr->nr_output_qs && \
> + ({ q = irq_ptr->output_qs[i]; 1; }); i++)

Hi Jose,

I generated a single commit of this part of your patch and applied
it to the s390 git tree.

Thanks!