All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v1] hw/i2c/aspeed_i2c: Fix Out-of-Bounds access by using dynamic register array
@ 2026-02-09  1:45 Jamin Lin
  2026-02-09  8:22 ` Cédric Le Goater
  0 siblings, 1 reply; 3+ messages in thread
From: Jamin Lin @ 2026-02-09  1:45 UTC (permalink / raw)
  To: Cédric Le Goater, Peter Maydell, Steven Lee, Troy Lee,
	Andrew Jeffery, Joel Stanley, open list:ASPEED BMCs,
	open list:All patches CC here
  Cc: Jamin Lin, Troy Lee, Kane Chen

The ASPEED I2C controller emulation used a fixed-size register array
(28 dwords) for all SoC variants, while multiple ASPEED SoCs
(AST2600, AST1030, AST2700) expose a larger MMIO register window
(e.g. reg_size = 0x80).

This mismatch allows MMIO accesses beyond the allocated register
array, leading to out-of-bounds reads in the I2C controller model.

Fix this by converting the register storage to a dynamically allocated
array sized according to the controller class reg_size. The register
array is now allocated during bus realize and free on unrealize,
ensuring safe access across different ASPEED SoC implementations.

This change eliminates I2C register out-of-bounds access caused by
SoC-specific register size differences.

Signed-off-by: Jamin Lin <jamin_lin@aspeedtech.com>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/3290
---
 include/hw/i2c/aspeed_i2c.h |  4 +---
 hw/i2c/aspeed_i2c.c         | 18 ++++++++++++++----
 2 files changed, 15 insertions(+), 7 deletions(-)

diff --git a/include/hw/i2c/aspeed_i2c.h b/include/hw/i2c/aspeed_i2c.h
index 68bd138026..205f0a58d2 100644
--- a/include/hw/i2c/aspeed_i2c.h
+++ b/include/hw/i2c/aspeed_i2c.h
@@ -36,8 +36,6 @@ OBJECT_DECLARE_TYPE(AspeedI2CState, AspeedI2CClass, ASPEED_I2C)
 #define ASPEED_I2C_NR_BUSSES 16
 #define ASPEED_I2C_SHARE_POOL_SIZE 0x800
 #define ASPEED_I2C_BUS_POOL_SIZE 0x20
-#define ASPEED_I2C_OLD_NUM_REG 11
-#define ASPEED_I2C_NEW_NUM_REG 28
 
 #define A_I2CD_M_STOP_CMD       BIT(5)
 #define A_I2CD_M_RX_CMD         BIT(3)
@@ -256,7 +254,7 @@ struct AspeedI2CBus {
     uint8_t id;
     qemu_irq irq;
 
-    uint32_t regs[ASPEED_I2C_NEW_NUM_REG];
+    uint32_t *regs;
     uint8_t pool[ASPEED_I2C_BUS_POOL_SIZE];
     uint64_t dma_dram_offset;
 };
diff --git a/hw/i2c/aspeed_i2c.c b/hw/i2c/aspeed_i2c.c
index fb3d6a5600..cf3a003978 100644
--- a/hw/i2c/aspeed_i2c.c
+++ b/hw/i2c/aspeed_i2c.c
@@ -1091,10 +1091,9 @@ static const MemoryRegionOps aspeed_i2c_bus_pool_ops = {
 
 static const VMStateDescription aspeed_i2c_bus_vmstate = {
     .name = TYPE_ASPEED_I2C,
-    .version_id = 6,
-    .minimum_version_id = 6,
+    .version_id = 7,
+    .minimum_version_id = 7,
     .fields = (const VMStateField[]) {
-        VMSTATE_UINT32_ARRAY(regs, AspeedI2CBus, ASPEED_I2C_NEW_NUM_REG),
         VMSTATE_UINT8_ARRAY(pool, AspeedI2CBus, ASPEED_I2C_BUS_POOL_SIZE),
         VMSTATE_UINT64(dma_dram_offset, AspeedI2CBus),
         VMSTATE_END_OF_LIST()
@@ -1465,8 +1464,9 @@ static const TypeInfo aspeed_i2c_bus_slave_info = {
 static void aspeed_i2c_bus_reset(DeviceState *dev)
 {
     AspeedI2CBus *s = ASPEED_I2C_BUS(dev);
+    AspeedI2CClass *aic = ASPEED_I2C_GET_CLASS(s->controller);
 
-    memset(s->regs, 0, sizeof(s->regs));
+    memset(s->regs, 0, aic->reg_size);
     i2c_end_transfer(s->bus);
 }
 
@@ -1492,6 +1492,7 @@ static void aspeed_i2c_bus_realize(DeviceState *dev, Error **errp)
     s->slave = i2c_slave_create_simple(s->bus, TYPE_ASPEED_I2C_BUS_SLAVE,
                                        0xff);
 
+    s->regs = g_new(uint32_t, aic->reg_size >> 2);
     memory_region_init_io(&s->mr, OBJECT(s), &aspeed_i2c_bus_ops,
                           s, s->name, aic->reg_size);
     sysbus_init_mmio(SYS_BUS_DEVICE(dev), &s->mr);
@@ -1501,6 +1502,14 @@ static void aspeed_i2c_bus_realize(DeviceState *dev, Error **errp)
     sysbus_init_mmio(SYS_BUS_DEVICE(dev), &s->mr_pool);
 }
 
+static void aspeed_i2c_bus_unrealize(DeviceState *dev)
+{
+    AspeedI2CBus *s = ASPEED_I2C_BUS(dev);
+
+    g_free(s->regs);
+    s->regs = NULL;
+}
+
 static const Property aspeed_i2c_bus_properties[] = {
     DEFINE_PROP_UINT8("bus-id", AspeedI2CBus, id, 0),
     DEFINE_PROP_LINK("controller", AspeedI2CBus, controller, TYPE_ASPEED_I2C,
@@ -1514,6 +1523,7 @@ static void aspeed_i2c_bus_class_init(ObjectClass *klass, const void *data)
 
     dc->desc = "Aspeed I2C Bus";
     dc->realize = aspeed_i2c_bus_realize;
+    dc->unrealize = aspeed_i2c_bus_unrealize;
     device_class_set_legacy_reset(dc, aspeed_i2c_bus_reset);
     device_class_set_props(dc, aspeed_i2c_bus_properties);
 }
-- 
2.43.0


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH v1] hw/i2c/aspeed_i2c: Fix Out-of-Bounds access by using dynamic register array
  2026-02-09  1:45 [PATCH v1] hw/i2c/aspeed_i2c: Fix Out-of-Bounds access by using dynamic register array Jamin Lin
@ 2026-02-09  8:22 ` Cédric Le Goater
  2026-02-10  2:23   ` Jamin Lin
  0 siblings, 1 reply; 3+ messages in thread
From: Cédric Le Goater @ 2026-02-09  8:22 UTC (permalink / raw)
  To: Jamin Lin, Peter Maydell, Steven Lee, Troy Lee, Andrew Jeffery,
	Joel Stanley, open list:ASPEED BMCs,
	open list:All patches CC here
  Cc: Troy Lee, Kane Chen

On 2/9/26 02:45, Jamin Lin wrote:
> The ASPEED I2C controller emulation used a fixed-size register array
> (28 dwords) for all SoC variants, while multiple ASPEED SoCs
> (AST2600, AST1030, AST2700) expose a larger MMIO register window
> (e.g. reg_size = 0x80).
> 
> This mismatch allows MMIO accesses beyond the allocated register
> array, leading to out-of-bounds reads in the I2C controller model.
> 
> Fix this by converting the register storage to a dynamically allocated
> array sized according to the controller class reg_size. The register
> array is now allocated during bus realize and free on unrealize,
> ensuring safe access across different ASPEED SoC implementations.
> 
> This change eliminates I2C register out-of-bounds access caused by
> SoC-specific register size differences.
> 
> Signed-off-by: Jamin Lin <jamin_lin@aspeedtech.com>
> Resolves: https://gitlab.com/qemu-project/qemu/-/issues/3290
> ---
>   include/hw/i2c/aspeed_i2c.h |  4 +---
>   hw/i2c/aspeed_i2c.c         | 18 ++++++++++++++----
>   2 files changed, 15 insertions(+), 7 deletions(-)
> 
> diff --git a/include/hw/i2c/aspeed_i2c.h b/include/hw/i2c/aspeed_i2c.h
> index 68bd138026..205f0a58d2 100644
> --- a/include/hw/i2c/aspeed_i2c.h
> +++ b/include/hw/i2c/aspeed_i2c.h
> @@ -36,8 +36,6 @@ OBJECT_DECLARE_TYPE(AspeedI2CState, AspeedI2CClass, ASPEED_I2C)
>   #define ASPEED_I2C_NR_BUSSES 16
>   #define ASPEED_I2C_SHARE_POOL_SIZE 0x800
>   #define ASPEED_I2C_BUS_POOL_SIZE 0x20
> -#define ASPEED_I2C_OLD_NUM_REG 11
> -#define ASPEED_I2C_NEW_NUM_REG 28


The ASPEED_I2C_NEW_NUM_REG value correctly covers the supported range
of registers of both 'old' and 'new' implementations.

Instead of dynamically allocating the regs array, which introduces
vmstate complexity, I would fix the way the array is accessed.

In aspeed_i2c_bus_old_read(), change :
    uint64_t value = bus->regs[offset / sizeof(*bus->regs)];
to
    uint64_t value = -1;

and adjust how 'value' is assigned else where.

Same for aspeed_i2c_bus_new_read().

If you want to keep the dynamic allocation, the vmstate needs a fix.
May be introduce a 'regs_nr' attribute for that and check how other
models save/load dynamic arrays.

Thanks,

C.



^ permalink raw reply	[flat|nested] 3+ messages in thread

* RE: [PATCH v1] hw/i2c/aspeed_i2c: Fix Out-of-Bounds access by using dynamic register array
  2026-02-09  8:22 ` Cédric Le Goater
@ 2026-02-10  2:23   ` Jamin Lin
  0 siblings, 0 replies; 3+ messages in thread
From: Jamin Lin @ 2026-02-10  2:23 UTC (permalink / raw)
  To: Cédric Le Goater, Peter Maydell, Steven Lee, Troy Lee,
	Andrew Jeffery, Joel Stanley, open list:ASPEED BMCs,
	open list:All patches CC here
  Cc: Troy Lee, Kane Chen

Hi Cédric,

> Subject: Re: [PATCH v1] hw/i2c/aspeed_i2c: Fix Out-of-Bounds access by using
> dynamic register array
> 
> On 2/9/26 02:45, Jamin Lin wrote:
> > The ASPEED I2C controller emulation used a fixed-size register array
> > (28 dwords) for all SoC variants, while multiple ASPEED SoCs (AST2600,
> > AST1030, AST2700) expose a larger MMIO register window (e.g. reg_size
> > = 0x80).
> >
> > This mismatch allows MMIO accesses beyond the allocated register
> > array, leading to out-of-bounds reads in the I2C controller model.
> >
> > Fix this by converting the register storage to a dynamically allocated
> > array sized according to the controller class reg_size. The register
> > array is now allocated during bus realize and free on unrealize,
> > ensuring safe access across different ASPEED SoC implementations.
> >
> > This change eliminates I2C register out-of-bounds access caused by
> > SoC-specific register size differences.
> >
> > Signed-off-by: Jamin Lin <jamin_lin@aspeedtech.com>
> > Resolves: https://gitlab.com/qemu-project/qemu/-/issues/3290
> > ---
> >   include/hw/i2c/aspeed_i2c.h |  4 +---
> >   hw/i2c/aspeed_i2c.c         | 18 ++++++++++++++----
> >   2 files changed, 15 insertions(+), 7 deletions(-)
> >
> > diff --git a/include/hw/i2c/aspeed_i2c.h b/include/hw/i2c/aspeed_i2c.h
> > index 68bd138026..205f0a58d2 100644
> > --- a/include/hw/i2c/aspeed_i2c.h
> > +++ b/include/hw/i2c/aspeed_i2c.h
> > @@ -36,8 +36,6 @@ OBJECT_DECLARE_TYPE(AspeedI2CState,
> AspeedI2CClass, ASPEED_I2C)
> >   #define ASPEED_I2C_NR_BUSSES 16
> >   #define ASPEED_I2C_SHARE_POOL_SIZE 0x800
> >   #define ASPEED_I2C_BUS_POOL_SIZE 0x20 -#define
> > ASPEED_I2C_OLD_NUM_REG 11 -#define ASPEED_I2C_NEW_NUM_REG 28
> 
> 
> The ASPEED_I2C_NEW_NUM_REG value correctly covers the supported range
> of registers of both 'old' and 'new' implementations.
> 
> Instead of dynamically allocating the regs array, which introduces
> vmstate complexity, I would fix the way the array is accessed.
> 
> In aspeed_i2c_bus_old_read(), change :
>     uint64_t value = bus->regs[offset / sizeof(*bus->regs)];
> to
>     uint64_t value = -1;
> 
> and adjust how 'value' is assigned else where.
> 
> Same for aspeed_i2c_bus_new_read().
> 
> If you want to keep the dynamic allocation, the vmstate needs a fix.
> May be introduce a 'regs_nr' attribute for that and check how other
> models save/load dynamic arrays.
> 
Thanks for the review and suggestions.
I will simply increase the size of the static array to fix this issue.

Thanks,
Jamin
> Thanks,
> 
> C.


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-02-10  2:24 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-02-09  1:45 [PATCH v1] hw/i2c/aspeed_i2c: Fix Out-of-Bounds access by using dynamic register array Jamin Lin
2026-02-09  8:22 ` Cédric Le Goater
2026-02-10  2:23   ` Jamin Lin

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.