search for: r535_gsp_retain_logging

.... void r535_gsp_dtor(struct nvkm_gsp *gsp) { @@ -2514,6 +2864,14 @@ r535_gsp_dtor(struct nvkm_gsp *gsp) nvkm_gsp_mem_dtor(&gsp->rmargs); nvkm_gsp_mem_dtor(&gsp->wpr_meta); nvkm_gsp_mem_dtor(&gsp->shm.mem); + +#ifdef CONFIG_DEBUG_FS + r535_gsp_retain_logging(gsp); + + kfree(gsp->blob_pmu.data); + gsp->blob_pmu.data = NULL; Please move this to r535_gsp_libos_debugfs_fini() to make it a bit more obvious why this needs to be cleaned up here. So first, I made a mistake, and the #ifdef is only supposed to be around the call to r535_...

...ose checks, please remove them. Ok > > + nvkm_debug(&gsp->subdev, "created debugfs GSP-RM logging entries\n"); > > + > > + if (keep_gsp_logging) { > > + nvkm_warn(&gsp->subdev, > > nvkm_info() is probably enough. Ok. > > +static void r535_gsp_retain_logging(struct nvkm_gsp *gsp) > > +{ > > + struct device *dev = gsp->subdev.device->dev; > > + struct dentry *root, *dir; > > + struct r535_gsp_log *log; > > + int ret; > > + > > + /* If we were told not to keep logs, then don't. */ > > + if (!keep_...

...> v01_00. > > So "previous RPC version (if it exists)" < "new RPC version" <= "current > VGPU version" at the time the RPC was created/updated. Pretty interesting, thanks for clarifying! (No need to document it though.) > > > +static void r535_gsp_retain_logging(struct nvkm_gsp *gsp) > > > +{ > > > + struct device *dev = gsp->subdev.device->dev; > > > + struct dentry *root, *dir; > > > + struct r535_gsp_log *log; > > > + int ret; > > > + > > > + /* If we were told not to keep logs, then d...

...size; > + dir = debugfs_create_blob(name, 0444, parent, d); > + if (IS_ERR(dir)) { > + kfree(d->data); > + memset(d, 0, sizeof(*d)); > + return PTR_ERR(dir); > + } > + > + i_size_write(d_inode(dir), d->size); > + > + return 0; > +} > + > +/** > + * r535_gsp_retain_logging - copy logging buffers to new debugfs root > + * @gsp: gsp pointer > + * > + * If keep_gsp_logging is enabled, then we want to preserve the GSP-RM logging > + * buffers and their debugfs entries, but all those objects would normally > + * deleted if GSP-RM fails to load. So we create...

...is all about debugfs, and since I don't want to blindly allocate buffers and linked lists that won't actually do anything, I would prefer that the code bails early if the infrastructure is not there. > > I tested this, and if drm_core_init() fails to create the dentry, then > > r535_gsp_retain_logging() will just keep going trying to create debugfs > > entries, because a root of NULL is actually valid, and the entries are > > created in /sys/kernel/debug/0000:65:00.0/ instead of > > /sys/kernel/debug/dri/0000:65:00.0/ > > This is because debugfs_lookup() doesn't retu...

...KERNEL); + if (!d->data) + return -ENOMEM; + + d->size = s->size; + dir = debugfs_create_blob(name, 0444, parent, d); + if (IS_ERR(dir)) { + kfree(d->data); + memset(d, 0, sizeof(*d)); + return PTR_ERR(dir); + } + + i_size_write(d_inode(dir), d->size); + + return 0; +} + +/** + * r535_gsp_retain_logging - copy logging buffers to new debugfs root + * @gsp: gsp pointer + * + * If keep_gsp_logging is enabled, then we want to preserve the GSP-RM logging + * buffers and their debugfs entries, but all those objects would normally + * deleted if GSP-RM fails to load. So we create a new debugfs root, allo...

Store the struct device pointer used to allocate the DMA buffer in the nvkm_gsp_mem object. This allows nvkm_gsp_mem_dtor() to release the buffer without needing the nvkm_gsp. This is needed so that we can retain DMA buffers even after the nvkm_gsp object is deleted. Signed-off-by: Timur Tabi <ttabi at nvidia.com> --- v2: rebased to drm-misc-next

Store the struct device pointer used to allocate the DMA buffer in the nvkm_gsp_mem object. This allows nvkm_gsp_mem_dtor() to release the buffer without needing the nvkm_gsp. This is needed so that we can retain DMA buffers even after the nvkm_gsp object is deleted. Signed-off-by: Timur Tabi <ttabi at nvidia.com> --- Notes: v2: added get/put_device calls

Store the struct device pointer used to allocate the DMA buffer in the nvkm_gsp_mem object. This allows nvkm_gsp_mem_dtor() to release the buffer without needing the nvkm_gsp. This is needed so that we can retain DMA buffers even after the nvkm_gsp object is deleted. Signed-off-by: Timur Tabi <ttabi at nvidia.com> v2: added get/put_device calls ---