[Bug] 给sdio的默认栈是否过小

RT-Thread Version

master

Affected area

Device drivers

Hardware/BSP vendor

HPMicro

Architecture

RISC-V

Board and hardware details

hpm63

Develop Toolchain

GCC

Describe the bug

默认的和idle相同的栈大小是不是有点太小了，在idle配成512时候，sdio挂网卡直接爆栈了。不知道其他情况会不会直接爆。

Other additional context

AI的分析：

当前链路是：

上下文 工作
━━━━━━━━━━━━━━━━━
SDXC 硬件 ISR 读取状态、屏蔽 DAT1、释放 host->sdio_irq_sem
─────────────────
sdio_irq 线程 加总线锁，通过 CMD52 读取 CCCR INT_PENDING，分发 function-1 回调，再使能 DAT1
─────────────────
ESP-Hosted 回调 只释放 port irq_sem
─────────────────
sdio_read 线程 读取 ESP 中断状态及实际数据

栈大的原因不是业务复杂，而是 CMD52 穿过了整套通用驱动调用链：

sdio_irq_thread
-> sdio_io_readb
-> sdio_io_rw_direct
-> mmcsd_send_cmd
-> mmcsd_send_request
-> hpm_sdmmc_request
-> hpm_sdmmc_transfer
-> hpm_sdmmc_transfer_interrupt_driven
-> rt_event_recv

仅编译器可见的函数栈帧就超过 624B。其中 hpm_sdmmc_request() 自己占 192B，因为栈上同时放了：

sdxc_adma_config_t
sdxc_xfer_t
sdxc_command_t
sdxc_data_t

再加锁、事件等待、线程入口和内核调用，实测就是 988B。当前还是 -O0，栈占用也会偏大。

Start with the SDIO IRQ thread and the HPMicro hpm_sdmmc_request() path described in the issue, then trace where the default SDIO thread stack is configured. Reproduce on the hpm63 board with GCC and idle configured to 512, measure stack use during SDIO network activity, and confirm that the default is large enough without causing overflow.