+ Fred, Chevron Li, Adrian
On Tue, 16 Jan 2024 at 23:18, Brent Roman <brent@xxxxxxxxx> wrote:
>
> Hi,
>
> I have an Intel Hades Canyon NUC (NUC8i7HVK)
> whose O2 Micro based SD-Card reader is no longer identified when its
> Linux kernel is updated past 5.12
>
> I "fixed" this by reverting a change from 5/9/21 (git
> efc58a96adcd29cc37487a60582d9d08b34f6640)
> that inserted proper error checking after all the PCI config space reads
> in the device probe.
> This would be code removed enclosed in #if 0 below:
>
> case PCI_DEVICE_ID_O2_SEABIRD0:
> case PCI_DEVICE_ID_O2_SEABIRD1:
> /* UnLock WP */
> ret = pci_read_config_byte(chip->pdev,
> O2_SD_LOCK_WP, &scratch);
> if (ret)
> return ret;
>
> scratch &= 0x7f;
> pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch);
>
> ret = pci_read_config_dword(chip->pdev,
> O2_SD_PLL_SETTING, &scratch_32);
> #if 0
> if (ret)
> return ret;
> #endif
>
> if ((scratch_32 & 0xff000000) == 0x01000000) {
> scratch_32 &= 0x0000FFFF;
> scratch_32 |= 0x1F340000;
>
> pci_write_config_dword(chip->pdev,
> O2_SD_PLL_SETTING, scratch_32);
> } else {
> scratch_32 &= 0x0000FFFF;
> scratch_32 |= 0x25100000;
>
> pci_write_config_dword(chip->pdev,
> O2_SD_PLL_SETTING, scratch_32);
>
> ret = pci_read_config_dword(chip->pdev,
> O2_SD_FUNC_REG4,
> &scratch_32);
> #if 0
> if (ret)
> return ret;
> #endif
> scratch_32 |= (1 << 22);
> pci_write_config_dword(chip->pdev,
> O2_SD_FUNC_REG4, scratch_32);
> }
>
> Both those pci_read_config_dword() calls return -EINVAL on my machine.
> The driver had been working earlier precisely because it was ignoring
> these error returns from pci_read_config_dword.
> Have you seen this behavior before on any other hardware?
>
> The SDcard reader identifies as:
> 03:00.0 SD Host controller: O2 Micro, Inc. SD/MMC Card Reader Controller
> (rev 01)
> 03:00.0 0805: 1217:8621 (rev 01)
>
> I've been unable to find /any/ information on this chip.
> Do you have any you could share? A datasheet would be ideal :-)
>
> Also:
> I've always had to operate this driver with debug_quirks2=4 to disable
> ultra high-speed support.
> Is this a known issue?
> Or, could it be a symptom of the failing pci_read_configs curing probe?
>
> Thanks for your attention,
Looks like the offending commit efc58a96adcd ("mmc: sdhci-pci-o2micro:
Add missing checks in sdhci_pci_o2_probe"), may not have been
thoroughly tested. Perhaps a revert is needed.
Let's see if Fred/Chevron Li has some thoughts around this.
Kind regards
Uffe
