| Message ID | 20240820123707.1788370-1-ruanjinjie@huawei.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [-next] spi: stm32-qspi: Simpify resource lookup | expand |
On Tue, Aug 20, 2024 at 08:37:07PM +0800, Jinjie Ruan wrote: > Use the devm_platform_ioremap_resource_byname() helper instead of > calling platform_get_resource_byname() and devm_ioremap_resource() > separately. This breaks boot on the Avenger96 board, it causes a NULL pointer dereference: [ 2.350480] Unable to handle kernel NULL pointer dereference at virtual address 00000000 when read ... [ 2.695787] Call trace: [ 2.695807] stm32_qspi_probe from platform_probe+0x5c/0xb0 [ 2.703914] platform_probe from really_probe+0xc8/0x2c8 [ 2.709284] really_probe from __driver_probe_device+0x88/0x19c [ 2.715145] __driver_probe_device from driver_probe_device+0x30/0x104 https://lava.sirena.org.uk/scheduler/job/650792 > - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi"); > - qspi->io_base = devm_ioremap_resource(dev, res); > + qspi->io_base = devm_platform_ioremap_resource_byname(pdev, "qspi"); > if (IS_ERR(qspi->io_base)) > return PTR_ERR(qspi->io_base); > > qspi->phys_base = res->start; > > - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi_mm"); > - qspi->mm_base = devm_ioremap_resource(dev, res); > + qspi->mm_base = devm_platform_ioremap_resource_byname(pdev, "qspi_mm"); > if (IS_ERR(qspi->mm_base)) > return PTR_ERR(qspi->mm_base); I can't identify any obvious error here, these look like a direct subsitution - the implementation of devm_platform_ioremap_resource_byname() looks to be the same as the replaced code and dev is set to &pdev->dev but I'm seeing the above behaviour.
On 2024/8/22 22:58, Mark Brown wrote: > On Tue, Aug 20, 2024 at 08:37:07PM +0800, Jinjie Ruan wrote: >> Use the devm_platform_ioremap_resource_byname() helper instead of >> calling platform_get_resource_byname() and devm_ioremap_resource() >> separately. > > This breaks boot on the Avenger96 board, it causes a NULL pointer > dereference: > > [ 2.350480] Unable to handle kernel NULL pointer dereference at virtual address 00000000 when read > > ... > > [ 2.695787] Call trace: > [ 2.695807] stm32_qspi_probe from platform_probe+0x5c/0xb0 > [ 2.703914] platform_probe from really_probe+0xc8/0x2c8 > [ 2.709284] really_probe from __driver_probe_device+0x88/0x19c > [ 2.715145] __driver_probe_device from driver_probe_device+0x30/0x104 > > https://lava.sirena.org.uk/scheduler/job/650792 > >> - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi"); >> - qspi->io_base = devm_ioremap_resource(dev, res); >> + qspi->io_base = devm_platform_ioremap_resource_byname(pdev, "qspi"); >> if (IS_ERR(qspi->io_base)) >> return PTR_ERR(qspi->io_base); >> >> qspi->phys_base = res->start; >> >> - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi_mm"); >> - qspi->mm_base = devm_ioremap_resource(dev, res); >> + qspi->mm_base = devm_platform_ioremap_resource_byname(pdev, "qspi_mm"); >> if (IS_ERR(qspi->mm_base)) >> return PTR_ERR(qspi->mm_base); > > I can't identify any obvious error here, these look like a direct > subsitution - the implementation of devm_platform_ioremap_resource_byname() > looks to be the same as the replaced code and dev is set to &pdev->dev > but I'm seeing the above behaviour. Mark, thank you very much, I can't identify the error here too. This bug may be easy to locate if have a reproduction environment, but it's good to keep it the way it is.
On 2024/8/22 22:58, Mark Brown wrote: > On Tue, Aug 20, 2024 at 08:37:07PM +0800, Jinjie Ruan wrote: >> Use the devm_platform_ioremap_resource_byname() helper instead of >> calling platform_get_resource_byname() and devm_ioremap_resource() >> separately. > > This breaks boot on the Avenger96 board, it causes a NULL pointer > dereference: > > [ 2.350480] Unable to handle kernel NULL pointer dereference at virtual address 00000000 when read > > ... > > [ 2.695787] Call trace: > [ 2.695807] stm32_qspi_probe from platform_probe+0x5c/0xb0 > [ 2.703914] platform_probe from really_probe+0xc8/0x2c8 > [ 2.709284] really_probe from __driver_probe_device+0x88/0x19c > [ 2.715145] __driver_probe_device from driver_probe_device+0x30/0x104 > > https://lava.sirena.org.uk/scheduler/job/650792 > >> - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi"); >> - qspi->io_base = devm_ioremap_resource(dev, res); >> + qspi->io_base = devm_platform_ioremap_resource_byname(pdev, "qspi"); >> if (IS_ERR(qspi->io_base)) >> return PTR_ERR(qspi->io_base); >> >> qspi->phys_base = res->start; >> >> - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi_mm"); >> - qspi->mm_base = devm_ioremap_resource(dev, res); >> + qspi->mm_base = devm_platform_ioremap_resource_byname(pdev, "qspi_mm"); >> if (IS_ERR(qspi->mm_base)) >> return PTR_ERR(qspi->mm_base); > > I can't identify any obvious error here, these look like a direct > subsitution - the implementation of devm_platform_ioremap_resource_byname() > looks to be the same as the replaced code and dev is set to &pdev->dev > but I'm seeing the above behaviour. Sorry, from lkp@intel.com，I identify that the "res" returned by platform_get_resource_byname() is used by "qspi->phys_base = res->start;" later，so the devm_platform_ioremap_resource_byname() is not inapplicable here.
diff --git a/drivers/spi/spi-stm32-qspi.c b/drivers/spi/spi-stm32-qspi.c index 955c920c4b63..46081df73cb0 100644 --- a/drivers/spi/spi-stm32-qspi.c +++ b/drivers/spi/spi-stm32-qspi.c @@ -780,15 +780,13 @@ static int stm32_qspi_probe(struct platform_device *pdev) qspi = spi_controller_get_devdata(ctrl); qspi->ctrl = ctrl; - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi"); - qspi->io_base = devm_ioremap_resource(dev, res); + qspi->io_base = devm_platform_ioremap_resource_byname(pdev, "qspi"); if (IS_ERR(qspi->io_base)) return PTR_ERR(qspi->io_base); qspi->phys_base = res->start; - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "qspi_mm"); - qspi->mm_base = devm_ioremap_resource(dev, res); + qspi->mm_base = devm_platform_ioremap_resource_byname(pdev, "qspi_mm"); if (IS_ERR(qspi->mm_base)) return PTR_ERR(qspi->mm_base);
Use the devm_platform_ioremap_resource_byname() helper instead of calling platform_get_resource_byname() and devm_ioremap_resource() separately. Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com> --- drivers/spi/spi-stm32-qspi.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)