No, it's not needed*.

I'm still waiting TI email...

EDIT: * AFAIU, for now. It might be needed in the future.

Thanks javispedro, good to know.

No ideas what we are doing wrong...
Strange to me that some tests running normally.

So you at least got some test running? I haven't even gotten that far.

Apologies if I missed the obvious here, but does this driver enable Jazelle, or just 3D acceleration?

./services_test
----------------------- Start -----------------------
Try calling PVRSRVConnect an invalid argument:
OK
Call PVRSRVConnect a valid argument:
OK
Try calling PVRSRVEnumerateDevices with invalid puiNumDevices:
OK
Get number of devices from PVRSRVEnumerateDevices:
OK
.... Reported 1 devices
.... Device Number | id
0001 | 0x0002
Attempt to acquire device 1:
OK
OK
eDeviceType = PVRSRV_DEVICE_ID_MBX1
byVersionMajor = 1
byVersionMinor = 3
Display Class API: enumerate devices
OK
OK
Display Class API: open device
Display Class API: enumerate display formats
OK
OK
Display Class API: enumerate display dimensions
OK
OK
Display Class API: get the system (primary) buffer
OK
Display Class API: get the sysphys addr from system buffer handle
OK
Display Class API: map sysphys addr to MBX
OK
Display Class API: get sync object for system surface
OK
Attempt to create command queue size 4096:
OK
Attempt to look inside command queue:
psQueueInfo->ui32ReadOffset = 0
psQueueInfo->ui32WriteOffset = 0
psQueueInfo->pvUserLinQueue = 0x4002c040
psQueueInfo->pvUserLinQueue[50] = 0
Attempt to create 2MB parameter buffer:
OK
-----Begin direct single blit: clear screen-----
Attempt to acquire SlavePort:
-----Completed direct single blit: clear screen-----
-----Begin multiple overlapping direct blits-----
-----Completed multiple overlapping direct blits-----
-----Begin multiple overlapping direct batch blits-----
-----Completed multiple overlapping direct batch blits-----
-----Begin multiple scheduled overlapping blits-----
-----Completed multiple scheduled overlapping blits-----
-----Begin slaveport write test (10485760 bytes) -----
-----Single write: Batch size = 220 bytes , 0.170000s , 60234.60kb/sec
-----Single write: Batch size = 100 bytes , 0.180000s , 56888.02kb/sec
-----Single write: Batch size = 40 bytes , 0.270000s , 37925.35kb/sec
-----Single write: Batch size = 20 bytes , 0.430000s , 23813.59kb/sec
-----Batch write: Batch size = 220 bytes , 0.170000s , 60233.34kb/sec
-----Batch write: Batch size = 100 bytes , 0.200000s , 51198.24kb/sec
-----Batch write: Batch size = 40 bytes , 0.280000s , 36570.17kb/sec
-----Batch write: Batch size = 20 bytes , 0.390000s , 26255.51kb/sec
----- slaveport write test finished-----
-----Begin memory bandwidth/alloc test (max size = 20971520 bytes, copy multiple = 10X)
-----Mem = 41 x 524288 bytes, (Alloc/Free/Copy) 0.02000s,0.04000s,2.71000s, 0.00049s/alloc, 0.00098s/free, 77461.25kb/sec
-----Mem = 81 x 262144 bytes, (Alloc/Free/Copy) 0.02000s,0.02000s,1.67000s, 0.00025s/alloc, 0.00025s/free, 124167.66kb/sec
-----Mem = 161 x 131072 bytes, (Alloc/Free/Copy) 0.06000s,0.05000s,1.63000s, 0.00037s/alloc, 0.00031s/free, 126429.45kb/sec
-----Mem = 321 x 65536 bytes, (Alloc/Free/Copy) 0.08000s,0.09000s,1.61000s, 0.00025s/alloc, 0.00028s/free, 127602.48kb/sec
-----Mem = 641 x 32768 bytes, (Alloc/Free/Copy) 0.12000s,0.19000s,1.60000s, 0.00019s/alloc, 0.00030s/free, 128200.00kb/sec
-----Mem = 1281 x 16384 bytes, (Alloc/Free/Copy) 0.40000s,0.58000s,1.61000s, 0.00031s/alloc, 0.00045s/free, 127304.35kb/sec
-----Mem = 2033 x 8192 bytes, (Alloc/Free/Copy) 0.62000s,1.17000s,1.28000s, 0.00030s/alloc, 0.00058s/free, 127062.50kb/sec
-----Mem = 2032 x 4096 bytes, (Alloc/Free/Copy) 0.69000s,1.17000s,0.65000s, 0.00034s/alloc, 0.00058s/free, 125046.15kb/sec
-----Mem = 4063 x 2048 bytes, (Alloc/Free/Copy) 0.94000s,1.40000s,0.65000s, 0.00023s/alloc, 0.00034s/free, 125015.38kb/sec
-----Mem = 8123 x 1024 bytes, (Alloc/Free/Copy) 1.60000s,1.60000s,0.65000s, 0.00020s/alloc, 0.00020s/free, 124969.23kb/sec
----- Memory bandwidth test finished (max mem = 21495808 bytes, address range = 008C0300 to 01CC1000) -----
Attempt to destroy parameter buffer:
OK
Attempt to destroy command queue:
OK
Display Class API: unmap system surface from MBX
Display Class API: close the device
PVRSRVReleaseMBXInfo:
OK
PVRSRVDisconnect:
Disconnecting client 242
OK
---------------------End loop 1---------------------

IMHO
The problem is warning while compiling omaplcd_linux.c and mmap.c about incompatible pointer type.

nokia800:/usr/local/bin# ./egl_test
X Error of failed request: BadWindow (invalid Window parameter)
Major opcode of failed request: 3 (X_GetWindowAttributes)
Resource id in failed request: 0x0
Serial number of failed request: 7
Current serial number in output stream: 8


services_test run OK, module load, mbxdaemon running, device chmod'ed 666

omaplcd_linux.c:
irqreturn_t vsync_isr(int irq, void *dev_id, struct pt_regs *regs)
should be
static irqreturn_t vsync_isr(int irq, void *dev_id)
Also in mmap.c
function define as PKV_OFFSET_STRUCT *PVRMMapRegisterArea
but return PKV_OFFSET_STRUCT

Why, 3D, of course (as the title says).