drm/msm: Set max segment size earlier

Fixes the following splat on a6xx gen2+ (a640, a650, a660 families),
a6xx gen1 has smaller GMU allocations so they fit under the default
64K max segment size.

   ------------[ cut here ]------------
   DMA-API: msm_dpu ae01000.display-controller: mapping sg segment longer than device claims to support [len=126976] [max=65536]
   WARNING: CPU: 5 PID: 9 at kernel/dma/debug.c:1160 debug_dma_map_sg+0x288/0x314
   Modules linked in:
   CPU: 5 PID: 9 Comm: kworker/u16:0 Not tainted 6.3.0-rc2-debug+ torvalds#629
   Hardware name: Google Villager (rev1+) with LTE (DT)
   Workqueue: events_unbound deferred_probe_work_func
   pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
   pc : debug_dma_map_sg+0x288/0x314
   lr : debug_dma_map_sg+0x288/0x314
   sp : ffffffc00809b560
   x29: ffffffc00809b560 x28: 0000000000000060 x27: 0000000000000000
   x26: 0000000000010000 x25: 0000000000000004 x24: 0000000000000004
   x23: ffffffffffffffff x22: ffffffdb31693cc0 x21: ffffff8080935800
   x20: ffffff8087417400 x19: ffffff8087a45010 x18: 0000000000000000
   x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000010000
   x14: 0000000000000001 x13: ffffffffffffffff x12: ffffffffffffffff
   x11: 0000000000000000 x10: 000000000000000a x9 : ffffffdb2ff05e14
   x8 : ffffffdb31275000 x7 : ffffffdb2ff08908 x6 : 0000000000000000
   x5 : 0000000000000001 x4 : ffffffdb2ff08a74 x3 : ffffffdb31275008
   x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffffff80803a9a80
   Call trace:
    debug_dma_map_sg+0x288/0x314
    __dma_map_sg_attrs+0x80/0xe4
    dma_map_sgtable+0x30/0x4c
    get_pages+0x1d4/0x1e4
    msm_gem_pin_pages_locked+0xbc/0xf8
    msm_gem_pin_vma_locked+0x58/0xa0
    msm_gem_get_and_pin_iova_range+0x98/0xac
    a6xx_gmu_memory_alloc+0x7c/0x128
    a6xx_gmu_init+0x16c/0x9b0
    a6xx_gpu_init+0x38c/0x3e4
    adreno_bind+0x214/0x264
    component_bind_all+0x128/0x1f8
    msm_drm_bind+0x2b8/0x608
    try_to_bring_up_aggregate_device+0x88/0x1a4
    __component_add+0xec/0x13c
    component_add+0x1c/0x28
    dp_display_probe+0x3f8/0x43c
    platform_probe+0x70/0xc4
    really_probe+0x148/0x280
    __driver_probe_device+0xc8/0xe0
    driver_probe_device+0x44/0x100
    __device_attach_driver+0x64/0xdc
    bus_for_each_drv+0xb0/0xd8
    __device_attach+0xd8/0x168
    device_initial_probe+0x1c/0x28
    bus_probe_device+0x44/0xb0
    deferred_probe_work_func+0xc8/0xe0
    process_one_work+0x2e0/0x488
    process_scheduled_works+0x4c/0x50
    worker_thread+0x218/0x274
    kthread+0xf0/0x100
    ret_from_fork+0x10/0x20
   irq event stamp: 293712
   hardirqs last  enabled at (293711): [<ffffffdb2ff0893c>] vprintk_emit+0x160/0x25c
   hardirqs last disabled at (293712): [<ffffffdb30b48130>] el1_dbg+0x24/0x80
   softirqs last  enabled at (279520): [<ffffffdb2fe10420>] __do_softirq+0x21c/0x4bc
   softirqs last disabled at (279515): [<ffffffdb2fe16708>] ____do_softirq+0x18/0x24
   ---[ end trace 0000000000000000 ]---

Signed-off-by: Rob Clark <[email protected]>
Fixes: db735fc ("drm/msm: Set dma maximum segment size for mdss")
Reviewed-by: Dmitry Baryshkov <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Author: robclark

drivers/gpu/drm/msm/msm_drv.c

@@ -459,13 +459,13 @@ static int msm_drm_init(struct device *dev, const struct drm_driver *drv)
 	if (ret)
 		goto err_cleanup_mode_config;
 
+	dma_set_max_seg_size(dev, UINT_MAX);
+
 	/* Bind all our sub-components: */
 	ret = component_bind_all(dev, ddev);
 	if (ret)
 		goto err_deinit_vram;
 
-	dma_set_max_seg_size(dev, UINT_MAX);
-
 	msm_gem_shrinker_init(ddev);
 
 	if (priv->kms_init) {