Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.kernel > #84303

Bug#1085425: linux-image-6.10.12-riscv64: dwmmc_starfive causes filesystem corruption on emmc

Cross-posted to 2 groups.

Show all headers | View raw

[Multipart message — attachments visible in raw view] - view raw

Hi,

On 2024-10-20 22:51, Aurelien Jarno wrote:
> Hi,
> 
> The swiotlb messages have been introduced by this upstream commits,
> which has been backported in various stable branches:
> 
> commit 8396c793ffdf28bb8aee7cfe0891080f8cab7890
> Author: Sam Protsenko <semen.protsenko@linaro.org>
> Date:   Wed Mar 6 17:20:52 2024 -0600
> 
>     mmc: dw_mmc: Fix IDMAC operation with pages bigger than 4K
> 
> You can try to build a kernel with it reverted to see if it also fixes
> the corruption issue.
> 
> In the meantime, I am working on a patch.

Nice! I saw your patch, and I'm going to test it in a day or two (I have 
an unrelated $dayjob). I'll let you know.

Previously on this thread...

> Hmm I am actually surprised that you are able to boot your system with
> so many corruption. As the characters seems duplicated and the emmc bus
> is a 8-bit one, it could be a timing issue between the emmc and the
> controller. One thing to try is lowering the speed by removing the
> highspeed capabilities from the device tree.

It worked! No more corruption. Lowering the speed doesn't have too much 
performance impact: using a simple dd test with bs=1M: before: 28MB/s, 
after: 20MB/s. The emmc chip is bought from orange pi, said to be 
capable of HS400, and it runs great as the boot drive for 
cfarm423.cfarm.net. So the throughput seems to be severely limited by 
CPU, and it's a small price to pay comparing to data corruption.

I do have another Pine64 Star64 8GB board. If you want to make a patch 
to send to upstream, I can test on Star64 as well, since 
cap-mmc-highspeed and mmc-hs200-1_8v are in jh7110-common.dtsi, and it 
will affect Star64 and Milk-V Mars too (but I don't have a Milk-V Mars 
board).

Lemme paste 3 pieces of kernel log here. mmcblk0 is the emmc here.
(unrelated: "Freeing initrd memory: 49556K" somehow takes more than 10 
seconds on this board)

both sd card and emmc are installed, vanilla device tree:

[   15.815390] dwmmc_starfive 16020000.mmc: IDMAC supports 32-bit 
address mode.
[   15.815463] dwmmc_starfive 16020000.mmc: Using internal DMA 
controller.
[   15.815484] dwmmc_starfive 16020000.mmc: Version ID is 290a
[   15.815624] dwmmc_starfive 16020000.mmc: DW MMC controller at irq 
38,32 bit host data width,32 deep fifo
[   15.815882] dwmmc_starfive 16020000.mmc: Got CD GPIO
[   16.021586] mmc_host mmc1: Bus speed (slot 0) = 45511111Hz (slot req 
400000Hz, actual 399220HZ div = 57)
...
[   16.396485] mmc_host mmc1: Bus speed (slot 0) = 45511111Hz (slot req 
50000000Hz, actual 45511111HZ div = 0)
[   16.396950] mmc1: new high speed SDXC card at address aaaa
[   16.444861] mmcblk1: mmc1:aaaa SR128 119 GiB
[   16.457041]  mmcblk1: p1
...
[   17.048470] dwmmc_starfive 16010000.mmc: IDMAC supports 32-bit 
address mode.
[   17.048537] dwmmc_starfive 16010000.mmc: Using internal DMA 
controller.
[   17.048559] dwmmc_starfive 16010000.mmc: Version ID is 290a
[   17.048706] dwmmc_starfive 16010000.mmc: DW MMC controller at irq 
62,32 bit host data width,32 deep fifo
[   17.049341] mmc_host mmc0: card is non-removable.
...
[   17.257588] mmc_host mmc0: Bus speed (slot 0) = 45511111Hz (slot req 
400000Hz, actual 399220HZ div = 57)
[   17.608337] mmc_host mmc0: Bus speed (slot 0) = 45511111Hz (slot req 
52000000Hz, actual 45511111HZ div = 0)
[   17.608424] mmc_host mmc0: Bus speed (slot 0) = 45511111Hz (slot req 
100000000Hz, actual 45511111HZ div = 0)
[   17.652239] mmc0: new HS200 MMC card at address 0001
[   17.654631] mmcblk0: mmc0:0001 A3A564 233 GiB
[   17.662989]  mmcblk0: p1
[   17.665030] mmcblk0boot0: mmc0:0001 A3A564 4.00 MiB
[   17.671474] mmcblk0boot1: mmc0:0001 A3A564 4.00 MiB
[   17.676643] mmcblk0rpmb: mmc0:0001 A3A564 16.0 MiB, chardev (245:0)
...


Removing cap-mmc-highspeed and mmc-hs200-1_8v from mmc@16010000.mmc:

[   16.108557] dwmmc_starfive 16020000.mmc: IDMAC supports 32-bit 
address mode.
[   16.122625] dwmmc_starfive 16020000.mmc: Using internal DMA 
controller.
[   16.135744] dwmmc_starfive 16020000.mmc: Version ID is 290a
[   16.139766] dwmmc_starfive 16020000.mmc: DW MMC controller at irq 
36,32 bit host data width,32 deep fifo
[   16.150320] dwmmc_starfive 16020000.mmc: Got CD GPIO
...
[   16.370098] mmc_host mmc1: Bus speed (slot 0) = 45511111Hz (slot req 
400000Hz, actual 399220HZ div = 57)
[   16.904984] mmc_host mmc1: Bus speed (slot 0) = 45511111Hz (slot req 
50000000Hz, actual 45511111HZ div = 0)
[   16.915598] mmc1: new high speed SDXC card at address aaaa
[   16.971392] mmcblk1: mmc1:aaaa SR128 119 GiB
[   17.100619]  mmcblk1: p1
...
[   17.683130] dwmmc_starfive 16010000.mmc: IDMAC supports 32-bit 
address mode.
[   17.690297] dwmmc_starfive 16010000.mmc: Using internal DMA 
controller.
[   17.696988] dwmmc_starfive 16010000.mmc: Version ID is 290a
[   17.702698] dwmmc_starfive 16010000.mmc: DW MMC controller at irq 
62,32 bit host data width,32 deep fifo
[   17.713205] mmc_host mmc0: card is non-removable.
...
[   17.926100] mmc_host mmc0: Bus speed (slot 0) = 45511111Hz (slot req 
400000Hz, actual 399220HZ div = 57)
[   18.245335] mmc_host mmc0: Bus speed (slot 0) = 45511111Hz (slot req 
25000000Hz, actual 22755555HZ div = 1)
[   18.255853] mmc0: new MMC card at address 0001
[   18.262920] mmcblk0: mmc0:0001 A3A564 233 GiB
[   18.276554]  mmcblk0: p1
[   18.281168] mmcblk0boot0: mmc0:0001 A3A564 4.00 MiB
[   18.292896] mmcblk0boot1: mmc0:0001 A3A564 4.00 MiB
[   18.303161] mmcblk0rpmb: mmc0:0001 A3A564 16.0 MiB, chardev (245:0)
[   20.076504] EXT4-fs (mmcblk0p1): orphan cleanup on readonly fs
[   20.086008] EXT4-fs (mmcblk0p1): mounted filesystem 
06521a7d-c765-47e2-9d76-95af2fd5f933 ro with ordered data mode. Quota 
mode: none.
...


However, removing only mmc-hs200-1_8v will cause I/O error, and the 
mmc_host driver seems to be confused:

[   15.848996] dwmmc_starfive 16020000.mmc: IDMAC supports 32-bit 
address mode.
[   15.849071] dwmmc_starfive 16020000.mmc: Using internal DMA 
controller.
[   15.849092] dwmmc_starfive 16020000.mmc: Version ID is 290a
[   15.849161] dwmmc_starfive 16020000.mmc: DW MMC controller at irq 
38,32 bit host data width,32 deep fifo
[   15.849422] dwmmc_starfive 16020000.mmc: Got CD GPIO
[   16.055595] mmc_host mmc1: Bus speed (slot 0) = 45511111Hz (slot req 
400000Hz, actual 399220HZ div = 57)
...
[   16.526561] mmc_host mmc1: Bus speed (slot 0) = 45511111Hz (slot req 
50000000Hz, actual 45511111HZ div = 0)
[   16.527095] mmc1: new high speed SDXC card at address aaaa
[   16.581133] mmcblk1: mmc1:aaaa SR128 119 GiB
[   16.745566]  mmcblk1: p1
...
[   17.058868] dwmmc_starfive 16010000.mmc: IDMAC supports 32-bit 
address mode.
[   17.058936] dwmmc_starfive 16010000.mmc: Using internal DMA 
controller.
[   17.058957] dwmmc_starfive 16010000.mmc: Version ID is 290a
[   17.059059] dwmmc_starfive 16010000.mmc: DW MMC controller at irq 
62,32 bit host data width,32 deep fifo
[   17.059913] mmc_host mmc0: card is non-removable.
...
[   17.267613] mmc_host mmc0: Bus speed (slot 0) = 45511111Hz (slot req 
400000Hz, actual 399220HZ div = 57)
[   17.586657] mmc_host mmc0: Bus speed (slot 0) = 45511111Hz (slot req 
52000000Hz, actual 45511111HZ div = 0)
[   17.587836] mmc0: new DDR MMC card at address 0001
[   17.590844] mmcblk0: mmc0:0001 A3A564 233 GiB
[   17.979638] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
400000Hz, actual 397669HZ div = 103)
[   18.254838] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
52000000Hz, actual 40959999HZ div = 1)
[   18.263718]  mmcblk0: p1
[   18.265785] mmcblk0boot0: mmc0:0001 A3A564 4.00 MiB
[   18.276811] mmcblk0boot1: mmc0:0001 A3A564 4.00 MiB
[   18.282619] mmcblk0rpmb: mmc0:0001 A3A564 16.0 MiB, chardev (245:0)
[   22.150792] EXT4-fs (mmcblk1p1): orphan cleanup on readonly fs
[   22.154444] EXT4-fs (mmcblk1p1): mounted filesystem 
1aed1397-d718-47ae-8529-5e03e17b1d21 ro with ordered data mode. Quota 
mode: none.
...
[   30.019636] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
400000Hz, actual 397669HZ div = 103)
[   30.283902] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
52000000Hz, actual 40959999HZ div = 1)
[   30.324910] I/O error, dev mmcblk0, sector 17016 op 0x0:(READ) flags 
0x80700 phys_seg 49 prio class 0
[   30.359866] I/O error, dev mmcblk0, sector 17144 op 0x0:(READ) flags 
0x80700 phys_seg 33 prio class 0
[   30.369855] I/O error, dev mmcblk0, sector 17192 op 0x0:(READ) flags 
0x80700 phys_seg 27 prio class 0
[   30.387459] I/O error, dev mmcblk0, sector 17272 op 0x0:(READ) flags 
0x80700 phys_seg 17 prio class 0
[   30.397539] I/O error, dev mmcblk0, sector 17320 op 0x0:(READ) flags 
0x80700 phys_seg 11 prio class 0
[   30.402496] I/O error, dev mmcblk0, sector 17336 op 0x0:(READ) flags 
0x80700 phys_seg 9 prio class 0
[   30.417071] I/O error, dev mmcblk0, sector 17384 op 0x0:(READ) flags 
0x80700 phys_seg 3 prio class 0
[   30.419111] I/O error, dev mmcblk0, sector 17392 op 0x0:(READ) flags 
0x80700 phys_seg 2 prio class 0
[   30.421018] I/O error, dev mmcblk0, sector 17400 op 0x0:(READ) flags 
0x80700 phys_seg 1 prio class 0
[   30.639584] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
400000Hz, actual 397669HZ div = 103)
[   30.911948] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
52000000Hz, actual 40959999HZ div = 1)
[   30.913021] I/O error, dev mmcblk0, sector 17192 op 0x0:(READ) flags 
0x0 phys_seg 1 prio class 0
[   30.917407] Buffer I/O error on dev mmcblk0p1, logical block 101, 
async page read
...
[  337.983701] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
400000Hz, actual 397669HZ div = 103)
[  338.256119] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
52000000Hz, actual 40959999HZ div = 1)
[  338.418167] I/O error, dev mmcblk0, sector 17144 op 0x0:(READ) flags 
0x80700 phys_seg 139 prio class 0
[  338.427760] I/O error, dev mmcblk0, sector 17192 op 0x0:(READ) flags 
0x80700 phys_seg 133 prio class 0
[  338.441890] I/O error, dev mmcblk0, sector 17264 op 0x0:(READ) flags 
0x80700 phys_seg 124 prio class 0
[  338.443428] I/O error, dev mmcblk0, sector 17272 op 0x0:(READ) flags 
0x80700 phys_seg 123 prio class 0
[  338.456114] I/O error, dev mmcblk0, sector 17336 op 0x0:(READ) flags 
0x80700 phys_seg 115 prio class 0
[  338.465384] I/O error, dev mmcblk0, sector 17368 op 0x0:(READ) flags 
0x80700 phys_seg 111 prio class 0
[  338.468482] I/O error, dev mmcblk0, sector 17384 op 0x0:(READ) flags 
0x80700 phys_seg 109 prio class 0
[  338.470030] I/O error, dev mmcblk0, sector 17392 op 0x0:(READ) flags 
0x80700 phys_seg 108 prio class 0
[  338.471551] I/O error, dev mmcblk0, sector 17400 op 0x0:(READ) flags 
0x80700 phys_seg 107 prio class 0
[  338.903702] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
400000Hz, actual 397669HZ div = 103)
[  339.162919] mmc_host mmc0: Bus speed (slot 0) = 81919999Hz (slot req 
52000000Hz, actual 40959999HZ div = 1)
[  339.163980] I/O error, dev mmcblk0, sector 17192 op 0x0:(READ) flags 
0x0 phys_seg 1 prio class 0
[  339.165270] Buffer I/O error on dev mmcblk0p1, logical block 101, 
async page read
[  339.167350] Buffer I/O error on dev mmcblk0p1, logical block 101, 
async page read
...




-- 
Jing Luo
About me: https://jing.rocks/about/
GPG Fingerprint: 4E09 8D19 00AA 3F72 1899 2614 09B3 316E 13A1 1EFC

Back to linux.debian.kernel | Previous | Next — Previous in thread | Find similar | Unroll thread

Thread

Bug#1085425: linux-image-6.10.12-riscv64: dwmmc_starfive causes filesystem corruption on emmc Jing Luo <jing@jing.rocks> - 2024-10-19 19:10 +0200
  Bug#1085425: linux-image-6.10.12-riscv64: dwmmc_starfive causes filesystem corruption on emmc Jing Luo <jing@jing.rocks> - 2024-10-20 17:00 +0200

csiph-web