Skip to contentSkip to conversation

dmaengine: ste_dma40: turn d40_base phy_chans into a flexible array

PATCH Applied Rosen Penev 4 participants 4 messages May 31 – Jun 8

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 dmaengine: ste_dma40: turn d40_base phy_chans into a flexible array

Convert the separately-offset phy_chans pointer to a C99 flexible array member at the end of struct d40_base, and switch the allocation to struct_size(). The log_chans and memcpy_chans slots continue to live in the same allocation immediately after phy_chans, indexed via base->log_chans. This removes the hand-rolled pointer fixup that recomputed phy_chans from base + ALIGN(sizeof(struct d40_base), 4).

The ALIGN(sizeof(struct d40_base), 4) requirement is met implicitly by the C compiler when using a flexible array member. With struct d40_chan phy_chans[] as the last member, the C standard guarantees sizeof(struct d40_base) includes trailing padding to satisfy the alignment of the flexible array element type (struct d40_chan). Since struct d40_chan contains members like spinlock_t, pointers, and struct dma_chan — all with alignment ≥ 4 — the compiler ensures sizeof(struct d40_base) is already a multiple of _Alignof(struct d40_chan) >= 4. The struct_size() macro then computes sizeof(struct d40_base) + sizeof(struct d40_chan) * num_phy_chans, so phy_chans[0] lands at a properly aligned offset without needing the manual ALIGN.

Assisted-by Claude:Opus-4.7

---
v3: add min() calls
v2: added ALIGN description

 drivers/dma/ste_dma40.c | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)
drivers/dma/ste_dma40.c +8 -6
index 9b803c0aec25..0d9ffa3e2663 100644
@@ -602,7 +602,6 @@ struct d40_base {
 	struct dma_device		  dma_both;
 	struct dma_device		  dma_slave;
 	struct dma_device		  dma_memcpy;
-	struct d40_chan			 *phy_chans;
 	struct d40_chan			 *log_chans;
 	struct d40_chan			**lookup_log_chans;
 	struct d40_chan			**lookup_phy_chans;
@@ -621,6 +620,7 @@ struct d40_base {
 	u32				 *regs_interrupt;
 	u16				  gcc_pwr_off_mask;
 	struct d40_gen_dmac		  gen_dmac;
+	struct d40_chan			 phy_chans[];
 };
 
 static struct device *chan2dev(struct d40_chan *d40c)
@@ -3128,6 +3128,7 @@ static int __init d40_hw_detect_init(struct platform_device *pdev,
 	struct clk *clk;
 	void __iomem *virtbase;
 	struct d40_base *base;
+	size_t alloc_size;
 	int num_log_chans;
 	int num_phy_chans;
 	int num_memcpy_chans;
@@ -3185,22 +3186,24 @@ static int __init d40_hw_detect_init(struct platform_device *pdev,
 	else
 		num_phy_chans = 4 * (readl(virtbase + D40_DREG_ICFG) & 0x7) + 4;
 
+	num_phy_chans = min(num_phy_chans, STEDMA40_MAX_PHYS);
+
 	/* The number of channels used for memcpy */
 	if (plat_data->num_of_memcpy_chans)
 		num_memcpy_chans = plat_data->num_of_memcpy_chans;
 	else
 		num_memcpy_chans = ARRAY_SIZE(dma40_memcpy_channels);
 
+	num_memcpy_chans = min(num_memcpy_chans, D40_MEMCPY_MAX_CHANS);
 	num_log_chans = num_phy_chans * D40_MAX_LOG_CHAN_PER_PHY;
 
 	dev_info(dev,
 		 "hardware rev: %d with %d physical and %d logical channels\n",
 		 rev, num_phy_chans, num_log_chans);
 
-	base = devm_kzalloc(dev,
-		ALIGN(sizeof(struct d40_base), 4) +
-		(num_phy_chans + num_log_chans + num_memcpy_chans) *
-		sizeof(struct d40_chan), GFP_KERNEL);
+	alloc_size = struct_size(base, phy_chans, num_phy_chans);
+	alloc_size += sizeof(*base->log_chans) * (num_log_chans + num_memcpy_chans);
+	base = devm_kzalloc(dev, alloc_size, GFP_KERNEL);
 
 	if (!base)
 		return -ENOMEM;
@@ -3213,7 +3216,6 @@ static int __init d40_hw_detect_init(struct platform_device *pdev,
 	base->virtbase = virtbase;
 	base->plat_data = plat_data;
 	base->dev = dev;
-	base->phy_chans = ((void *)base) + ALIGN(sizeof(struct d40_base), 4);
 	base->log_chans = &base->phy_chans[num_phy_chans];
 
 	if (base->plat_data->num_of_phy_chans == 14) {

--
2.54.0

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

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

Pre-existing issues:
- [High] Out-of-bounds read and write on dma40_memcpy_channels due to incorrect array size and clamping limit. --

commit f80ed233c1567c0d1ee4933ba428492ac35dc443
Author: Rosen Penev <[email protected]>

dmaengine: ste_dma40: turn d40_base phy_chans into a flexible array

Convert the separately-offset phy_chans pointer to a C99 flexible array member at the end of struct d40_base, and switch the allocation to struct_size().

Quoted 4 lines
drivers/dma/ste_dma40.c +0 -0
index 9b803c0aec254..0d9ffa3e26639 100644

[ ... ]

drivers/dma/ste_dma40.c static int __init d40_hw_detect_init(struct platform_device *pdev, · line 3197

Quoted 13 lines
drivers/dma/ste_dma40.c +3 -0
@@ -3185,22 +3186,24 @@ static int __init d40_hw_detect_init(struct platform_device *pdev,
 	else
 		num_phy_chans = 4 * (readl(virtbase + D40_DREG_ICFG) & 0x7) + 4;
 
+	num_phy_chans = min(num_phy_chans, STEDMA40_MAX_PHYS);
+
 	/* The number of channels used for memcpy */
 	if (plat_data->num_of_memcpy_chans)
 		num_memcpy_chans = plat_data->num_of_memcpy_chans;
 	else
 		num_memcpy_chans = ARRAY_SIZE(dma40_memcpy_channels);
 
+	num_memcpy_chans = min(num_memcpy_chans, D40_MEMCPY_MAX_CHANS);

[Severity: High]
This isn't a bug introduced by this patch, but does this clamping limit allow out-of-bounds accesses on the dma40_memcpy_channels array?

The dma40_memcpy_channels array has exactly 6 elements, but D40_MEMCPY_MAX_CHANS is 8.

When d40_of_probe() is called, it permits num_memcpy up to 8 and writes them into the array:

drivers/dma/ste_dma40.c:d40_of_probe() {
    ...
    of_property_read_u32_array(np, "memcpy-channels",
                               dma40_memcpy_channels,
                               num_memcpy);
    ...
}

Can this cause an out-of-bounds write corrupting adjacent memory?

Additionally, because num_memcpy_chans is clamped to 8 here in d40_hw_detect_init(), it allows up to 8 memcpy channels to be registered.

When clients allocate these channels, d40_config_memcpy() indexes the array:

drivers/dma/ste_dma40.c:d40_config_memcpy() {
    ...
    d40c->dma_cfg.dev_type = dma40_memcpy_channels[d40c->chan.chan_id];
    ...
}

If chan_id is 6 or 7, won't this result in an out-of-bounds read?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
maintainer Reviewed-by on v3 1/1
#2 lore raw
In reply to Rosen Penev
Quoted 19 lines · Rosen Penev

Convert the separately-offset phy_chans pointer to a C99 flexible array member at the end of struct d40_base, and switch the allocation to struct_size(). The log_chans and memcpy_chans slots continue to live in the same allocation immediately after phy_chans, indexed via base->log_chans. This removes the hand-rolled pointer fixup that recomputed phy_chans from base + ALIGN(sizeof(struct d40_base), 4).

The ALIGN(sizeof(struct d40_base), 4) requirement is met implicitly by the C compiler when using a flexible array member. With struct d40_chan phy_chans[] as the last member, the C standard guarantees sizeof(struct d40_base) includes trailing padding to satisfy the alignment of the flexible array element type (struct d40_chan). Since struct d40_chan contains members like spinlock_t, pointers, and struct dma_chan — all with alignment ≥ 4 — the compiler ensures sizeof(struct d40_base) is already a multiple of _Alignof(struct d40_chan) >= 4. The struct_size() macro then computes sizeof(struct d40_base) + sizeof(struct d40_chan) * num_phy_chans, so phy_chans[0] lands at a properly aligned offset without needing the manual ALIGN.

Quoted 2 lines
Assisted-by Claude:Opus-4.7

Yours,
Linus Walleij

maintainer Applied
#3 lore raw
In reply to Rosen Penev
Quoted 8 lines · Rosen Penev

Convert the separately-offset phy_chans pointer to a C99 flexible array member at the end of struct d40_base, and switch the allocation to struct_size(). The log_chans and memcpy_chans slots continue to live in the same allocation immediately after phy_chans, indexed via base->log_chans. This removes the hand-rolled pointer fixup that recomputed phy_chans from base + ALIGN(sizeof(struct d40_base), 4).

[...]

Applied, thanks!

[1/1] dmaengine: ste_dma40: turn d40_base phy_chans into a flexible array

      commit: cc4fea19daeb0460fe3569e0a2d523f427b2bac1

Best regards,

Signature · 1 line
~Vinod