Hi Alexander, On Thu, Dec 07, 2023 at 08:52:46AM +0100, Alexander Stein wrote: > Am Mittwoch, 6. Dezember 2023, 23:36:36 CET schrieb Laurent Pinchart: > > On Wed, Dec 06, 2023 at 10:44:01AM +0100, Alexander Stein wrote: > > > The extended address registers are missing in the debug output register > > > list. Add them. > > > > > > Signed-off-by: Alexander Stein <alexander.stein@xxxxxxxxxxxxxxx> > > > --- > > > > > > drivers/media/platform/nxp/imx8-isi/imx8-isi-debug.c | 7 +++++++ > > > 1 file changed, 7 insertions(+) > > > > > > diff --git a/drivers/media/platform/nxp/imx8-isi/imx8-isi-debug.c > > > b/drivers/media/platform/nxp/imx8-isi/imx8-isi-debug.c index > > > 6709ab7ea1f3..3ac5685d6cc1 100644 > > > --- a/drivers/media/platform/nxp/imx8-isi/imx8-isi-debug.c > > > +++ b/drivers/media/platform/nxp/imx8-isi/imx8-isi-debug.c > > > @@ -66,6 +66,13 @@ static int mxc_isi_debug_dump_regs_show(struct seq_file *m, void *p) > > > MXC_ISI_DEBUG_REG(CHNL_OUT_BUF2_ADDR_V), > > > MXC_ISI_DEBUG_REG(CHNL_SCL_IMG_CFG), > > > MXC_ISI_DEBUG_REG(CHNL_FLOW_CTRL), > > > > > > + MXC_ISI_DEBUG_REG(CHNL_Y_BUF1_XTND_ADDR), > > > + MXC_ISI_DEBUG_REG(CHNL_U_BUF1_XTND_ADDR), > > > + MXC_ISI_DEBUG_REG(CHNL_V_BUF1_XTND_ADDR), > > > + MXC_ISI_DEBUG_REG(CHNL_Y_BUF2_XTND_ADDR), > > > + MXC_ISI_DEBUG_REG(CHNL_U_BUF2_XTND_ADDR), > > > + MXC_ISI_DEBUG_REG(CHNL_V_BUF2_XTND_ADDR), > > > + MXC_ISI_DEBUG_REG(CHNL_IN_BUF_XTND_ADDR), > > > > Those registers only exist in the i.MX8MP. The i.MX8MN referenece manual > > lists them as reserved and indicates they read as 0's, so it should be > > safe. The i.MX93 reference manual, however, doesn't list those > > registers at all, so accessing them may lead to issues. > > > > How what platform(s) have you tested this patch ? > > Ah, you are right. That's something I didn't take into account. Up until now I > only tested this on i.MX8MP. I hope I can test this on i.MX8MN and i.MX93 > soon. > Otherwise dumping these registers can depend on pdata->has_36bit_dma just to > be safe. That sounds like a good solution to me. Will you do so in a v2 ? > > > }; > > > > > > struct mxc_isi_pipe *pipe = m->private; -- Regards, Laurent Pinchart
