| Message ID | 201501301523.34410@pali (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Friday 30 January 2015 15:23:34 Pali Rohár wrote: > On Tuesday 27 January 2015 10:17:39 Pali Rohár wrote: > > On Tuesday 27 January 2015 00:21:04 Aaro Koskinen wrote: > > > Hi, > > > > > > On Mon, Jan 26, 2015 at 10:04:59PM +0100, Pali Rohár wrote: > > > > problem? Why any of these two patches fix problem when > > > > mmc is not detected by kernel in qemu (machine n900)? > > > > Detection of mmc fails because function > > > > mmc_send_op_cond() without one of above patches fails. > > > > > > Has it ever worked? It could be just that QEMU's emulation > > > is broken. Since the kernel works on actual HW, you > > > probably should contact QEMU maintainers. > > > > > > I don't see n900 in Debian's QEMU. There's n800 and n810 > > > but I couldn't boot any of my kernels with those... > > > > > > A. > > > > It worked with stock Nokia kernel (2.6.28). > > > > N900 qemu support is not in upstream qemu project, but in > > linaro qemu version. You can find it e.g. in ubuntu: > > http://packages.ubuntu.com/source/precise-updates/qemu-linar > > o > > Here is another alternative patch which enable internal eMMC > support in qemu n900: > > diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c > index 9584bff..225df49 100644 > --- a/drivers/mmc/core/core.c > +++ b/drivers/mmc/core/core.c > @@ -2330,13 +2330,13 @@ static int mmc_rescan_try_freq(struct > mmc_host *host, unsigned freq) > sdio_reset(host); > mmc_go_idle(host); > > - mmc_send_if_cond(host, host->ocr_avail); > - > - /* Order's important: probe SDIO, then SD, then MMC */ > - if (!mmc_attach_sdio(host)) > - return 0; > - if (!mmc_attach_sd(host)) > - return 0; > +// mmc_send_if_cond(host, host->ocr_avail); > +// > +// /* Order's important: probe SDIO, then SD, then MMC */ > +// if (!mmc_attach_sdio(host)) > +// return 0; > +// if (!mmc_attach_sd(host)) > +// return 0; > if (!mmc_attach_mmc(host)) > return 0; > > It just disable SDIO and SD code. It looks like using some > SDIO or SD commands on mmc controller in qemu cause > problems... > > Any idea? Pavel, see this email thread: http://www.spinics.net/lists/linux-mmc/msg30424.html You had problem with mounting mmc device as rootfs on (real) Nokia N900. Can you test one of those 3 patches which I sent to this email thread? If it change something for your kernel or userspace? I would like to know if I found another bug in qemu or same as you had...
On Friday 30 January 2015 15:23:34 Pali Rohár wrote: > On Tuesday 27 January 2015 10:17:39 Pali Rohár wrote: > > On Tuesday 27 January 2015 00:21:04 Aaro Koskinen wrote: > > > Hi, > > > > > > On Mon, Jan 26, 2015 at 10:04:59PM +0100, Pali Rohár wrote: > > > > problem? Why any of these two patches fix problem when mmc > > > > is not detected by kernel in qemu (machine n900)? > > > > Detection of mmc fails because function > > > > mmc_send_op_cond() without one of above patches fails. > > > > > > Has it ever worked? It could be just that QEMU's emulation > > > is broken. Since the kernel works on actual HW, you > > > probably should contact QEMU maintainers. > > > > > > I don't see n900 in Debian's QEMU. There's n800 and n810 but > > > I couldn't boot any of my kernels with those... > > > > > > A. > > > > It worked with stock Nokia kernel (2.6.28). > > > > N900 qemu support is not in upstream qemu project, but in > > linaro qemu version. You can find it e.g. in ubuntu: > > http://packages.ubuntu.com/source/precise-updates/qemu-linaro > > Here is another alternative patch which enable internal eMMC > support in qemu n900: > > diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c > index 9584bff..225df49 100644 > --- a/drivers/mmc/core/core.c > +++ b/drivers/mmc/core/core.c > @@ -2330,13 +2330,13 @@ static int mmc_rescan_try_freq(struct > mmc_host *host, unsigned freq) > sdio_reset(host); > mmc_go_idle(host); > > - mmc_send_if_cond(host, host->ocr_avail); > - > - /* Order's important: probe SDIO, then SD, then MMC */ > - if (!mmc_attach_sdio(host)) > - return 0; > - if (!mmc_attach_sd(host)) > - return 0; > +// mmc_send_if_cond(host, host->ocr_avail); > +// > +// /* Order's important: probe SDIO, then SD, then MMC */ > +// if (!mmc_attach_sdio(host)) > +// return 0; > +// if (!mmc_attach_sd(host)) > +// return 0; > if (!mmc_attach_mmc(host)) > return 0; > > It just disable SDIO and SD code. It looks like using some SDIO > or SD commands on mmc controller in qemu cause problems... > > Any idea? Now there are DT flags no-sdio and no-sd which skips mmc_attach_sdio() and mmc_attach_sd() calls. Same result as my commented code above... I will send patch for Nokia N900 which will disable SDIO and SD.
diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c index 9584bff..225df49 100644 --- a/drivers/mmc/core/core.c +++ b/drivers/mmc/core/core.c @@ -2330,13 +2330,13 @@ static int mmc_rescan_try_freq(struct mmc_host *host, unsigned freq) sdio_reset(host); mmc_go_idle(host); - mmc_send_if_cond(host, host->ocr_avail); - - /* Order's important: probe SDIO, then SD, then MMC */ - if (!mmc_attach_sdio(host)) - return 0; - if (!mmc_attach_sd(host)) - return 0; +// mmc_send_if_cond(host, host->ocr_avail); +// +// /* Order's important: probe SDIO, then SD, then MMC */ +// if (!mmc_attach_sdio(host)) +// return 0; +// if (!mmc_attach_sd(host)) +// return 0; if (!mmc_attach_mmc(host)) return 0;