I/O errors from SD card slot? (Manjaro, kernel 5.7)

[zackw]

I'm getting I/O errors from some, but not all, attempts to use the SD card slot in my PBP.  `badblocks -w /dev/mmcblk1` completes without error, and `fdisk` is also happy, but when I attempt to run `mkfs.ext4` or `cryptsetup luksFormat` on the (single, whole-disk) partition, the operation fails and I see errors like this in dmesg:

Code:

blk_update_request: I/O error, dev mmcblk1, sector 2120 op 0x1:(WRITE) flags 0x8800 phys_seg 255 prio class 0
blk_update_request: I/O error, dev mmcblk1, sector 2128 op 0x1:(WRITE) flags 0x8800 phys_seg 254 prio class 0
blk_update_request: I/O error, dev mmcblk1, sector 2117 op 0x1:(WRITE) flags 0x8800 phys_seg 255 prio class 0
blk_update_request: I/O error, dev mmcblk1, sector 2127 op 0x1:(WRITE) flags 0x8800 phys_seg 255 prio class 0

(Only one error occurred per attempt to format the partition.)

I put the card into a different computer and was able to create a filesystem successfully, so I don't think the card itself is at fault. However, when I put the formatted card back into the PBP and tried to mount the filesystem or to run e2fsck on it, I got more of the same errors, just with different sector numbers and flags:

Code:

blk_update_request: I/O error, dev mmcblk1, sector 8301 op 0x1:(WRITE) flags 0x8800 phys_seg 251 prio class 0
blk_update_request: I/O error, dev mmcblk1, sector 40960 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
blk_update_request: I/O error, dev mmcblk1, sector 59023360 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0

This is a 2019-batch PBP, so it shipped with MrFixit's modified Debian, and the card slot was working fine with that kernel; I only started seeing problems after I reflashed the internal storage with Manjaro. So right now I'm inclined to suspect a kernel bug. Anyone else seeing similar problems? Workarounds? Debugging advice?