[Bug] RT-Thread DM365 SPI divides by an unchecked zero clock rate
维护者通常 1 天内回复
评估
- 难度
- 2/5
- 预计耗时
- 1-3 小时
- 新手友好度
- 68/100
- Issue 类型
- 缺陷
- 描述清晰度
- 描述清楚
- 活跃度
- 活跃
- 技术栈
- c
- 领域
- embedded-iot
调研方向
阅读 bsp/dm365/drivers/spi-davinci.c 中 davinci_spi_get_prescale() 和 davinci_spi_setup_transfer() 附近的代码，然后追踪 components/drivers/spi/dev_spi_core.c 中的配置调用。添加一个速率为零的配置用例，并在任何传输发生前覆盖受支持的速度限制。完成的标准是：当配置发生更改且 max_hz == 0 时，在除法之前拒绝该配置，同时有效速率仍能成功配置。
由索引模型根据 Issue 内容生成。
描述
RT-Thread Version
v5.3.0
Affected area
Other / Not sure
Hardware/BSP vendor
Not applicable / Other
Architecture
Not applicable / Other
Board and hardware details
DM365 SPI
Develop Toolchain
Other
Describe the bug
RT-Thread DM365 SPI divides by an unchecked zero clock rate
Affected version
RT-Thread v5.3.0.
Describe the bug
davinci_spi_get_prescale() divides by the requested SPI frequency without rejecting zero. rt_spi_configure() copies max_hz into the device configuration and asks the bus to apply it.
An unchanged configuration returns before that call. A request that changes another field, such as data width 9, with max_hz == 0 does not. rt_spi_bus_configure() calls the driver only when this device currently owns the bus.
davinci_spi_setup_transfer() accepts data widths 2 through 16, copies cfg->max_hz, and calls the prescaler:
#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
ret = DIV_ROUND_UP(clk_get_rate(dspi->clk), max_speed_hz);
The prescaler clamps the result only after that division. A zero frequency is not turned into a default speed on this path.
Locations:
bsp/dm365/drivers/spi-davinci.c:35
bsp/dm365/drivers/spi-davinci.c:267
bsp/dm365/drivers/spi-davinci.c:272
bsp/dm365/drivers/spi-davinci.c:298
bsp/dm365/drivers/spi-davinci.c:313
bsp/dm365/drivers/spi-davinci.c:333
bsp/dm365/drivers/spi-davinci.c:788
components/drivers/spi/dev_spi_core.c:211
components/drivers/spi/dev_spi_core.c:277
components/drivers/spi/dev_spi_core.c:288
Threat model
The boundary is a caller of rt_spi_configure(device, cfg). This report does not treat a network peer or a byte on the SPI bus as the attacker. The value that reaches the division is cfg->max_hz. The caller also supplies cfg->mode and cfg->data_width. The clock rate returned for the controller is not caller input.
max_hz is a rt_uint32_t. Zero is the unsafe value. The core returns before calling the driver when mode, data width, and max_hz are all unchanged, so a request that only repeats the current configuration does not divide. A request that changes another field, such as setting data_width to 9 while leaving max_hz at 0, does reach the driver. data_width must also be 2 through 16 or the driver returns before the division. rt_spi_bus_configure() calls the driver only when this device currently owns the bus.
There is no shell command in this path. The entry is whatever product code can submit an SPI configuration: a trusted driver caller, or an untrusted service that forwards a frequency into rt_spi_configure().
Steps to reproduce
I have not reproduced this on DM365 hardware. The reduced check keeps the round-up macro and a positive clock.
#include <stdint.h>
#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
static int prescale(uint32_t max_hz)
{
return DIV_ROUND_UP(48000000u, max_hz);
}
int main(void)
{
return prescale(0);
}
Compile it with an undefined-behavior sanitizer, for example:
clang -fsanitize=undefined -g -O1 repro.c -o repro && ./repro
Returning an error when max_hz == 0, before the division, avoids the report. Positive rates such as 500 kHz complete the division.
Relevant log output
runtime error: division by zero
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior
Impact
A caller that can submit an SPI configuration for a DM365 SPI device, with a supported data width and a zero max_hz, reaches a division by zero in the driver when that device owns the bus. The reduced check shows the arithmetic fault. It does not show the processor exception on DM365 hardware, and this API is not by itself a network input.
Suggested fix
Reject zero before calling the prescaler, and keep a zero check inside the helper. If zero is intended to mean a default speed, resolve it to an explicit frequency first. Test a changed configuration whose max_hz is zero, plus the supported speed limits, before any transfer.
Other additional context
No response
- 主要语言
- C
- 星标
- 12.3k
- 派生
- 5.5k
- 平均合并
- 4 天 9 小时
- 30 天内合并 PR
- 38
环境准备
在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
RT-Thread/rt-thread 的其他 Issue
-
bug
难度 2/5 1-3 小时 新手友好度 67/100
RT-Thread/rt-thread#11951 · 1 条评论 ·
维护者通常 1 天内回复
-
bug
难度 2/5 1-3 小时 新手友好度 72/100
RT-Thread/rt-thread#11949 · 1 条评论 ·
维护者通常 1 天内回复
-
[Bug] RT-Thread AB32 ADC uses an unchecked channel as a register index可能已有人在做 @usualtreee 于 3 天前认领。 未关闭bug Component component: drivers
难度 2/5 1-3 小时 新手友好度 72/100
RT-Thread/rt-thread#11948 · 1 条评论 ·
维护者通常 1 天内回复
-
[Bug] [libc] timer_gettime 拆秒前将总纳秒截断为 uint32_t，剩余时间查询错误可能已有人在做 @Yunlk 于 10 天前认领。 未关闭in progress
难度 2/5 1-3 小时 新手友好度 68/100
RT-Thread/rt-thread#11924 · 7 条评论 · 已指派 1 人 ·
维护者通常 1 天内回复
-
[Bug] dfs_v2 tmpfs: ftruncate(fd, 0) frees the file buffer but reports success — use-after-free read/write and double free from a legal POSIX call可能已有人在做 @remukisama 于 14 天前认领。 未关闭in progress
难度 2/5 1-3 小时 新手友好度 78/100
RT-Thread/rt-thread#11902 · 4 条评论 · 已指派 1 人 ·
维护者通常 1 天内回复
查看 RT-Thread/rt-thread 的全部 Issue
相似的 Issue
-
Bug
难度 2/5 1-3 小时 新手友好度 66/100
维护者通常 1 天内回复
-
Integer overflow → heap out-of-bounds write in `contrib/pngminus/pnm2png.c` via crafted PNM width未关闭
难度 2/5 1-3 小时 新手友好度 78/100
维护者通常 4 天内回复
-
难度 2/5 1-3 小时 新手友好度 70/100
meshcore-dev/MeshCore#3760 · 1 条评论 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 68/100
ly05010419/MacNativeGames#1697 ·
维护者通常 1 天内回复
-
severity: low
难度 2/5 1-3 小时 新手友好度 72/100
luainkernel/lunatik#2318 ·
维护者通常 1 天内回复