PS3 linux patch update and issue

[Sombat Tragoongoson]

Hello,

I find some issue for latest kernel update.
see my note.
hope this help.

 

[Sombat Tragoongoson]

Hello,

Update linux-5.15 patches and my test app. for to test.
I give the test source(vertex light and fragment light).

still hope to see the porting home brew game.

Have fun.

 

[Sombat Tragoongoson]

Hello,

some may be looking for latest linux kernel for ps3 otheros++.

two kernel 6.8.1 and 6.9-rc1 testing on my slim(355).

they are still boot fine for latest Gentoo and T2 SDE.

but there are some issues.

1) config for Build big-endian kernel using ELF ABI V2
kernel 6.8.1 cannot boot.
kernel 6.9-rc1 can boot OK
know not much about this config for.

2) kernel 6.9-rc1
ps3disk.c: ps3disk_probe line 379

struct queue_limits lim = {
.logical_block_size = dev->blk_size,
.max_hw_sectors = dev->bounce_size >> 9,
.max_segments = -1,
.max_segment_size = dev->bounce_size,
.dma_alignment = dev->blk_size - 1,
};

the temporlary variable lim, it seems proper,
but it not working for me. some other members of this struct
may not be zero. they are using for deault settings.
fixed by memset and set its members as above.

3) for multi-block queues
this below just my understanding if not right,
please do help advice.

hardware request queues can run all,
but for ps3, it handdles one at a time
for each reuest to read, write or flush.
it will then cause one interrupt
to tell the result whether successful or not.

in the function(ps3disk_queue_rq),
if lv1_storage_write, lv1_storage_read or
lv1_storage_send_device_command is result OK.
then priv->req = req. this will be set to NULL when interrupt.

for next two or more requsts, it will check this
if not NULL, it will return BLK_STS_DEV_RESOURCE.
then this request will re-run by
blk_mq_dispatch_rq_list(blk-mq.c)
this will hold control one at a time.

Hope this will help for some.

 

[Sombat Tragoongoson]

Hello,

just take notice,
ps3linux_kernel_patches_6_8.tar.gz is not proper.
it is 6.9. really sorry for that mistaken.
so need new update one.

another, about booting issue for multi-block queue.

1) Gentoo, no any problems.

2) T2-SDE is fit for "mq_dead_line" schedualer (default)
and this can switch to "bfq" after boot.

if using "bfq",
the udev discovery seems to be time-out and then kill the job.
especially for the first three or more such as ps3disk(a,b and .).
then the result can not do reboot.
just try pressing power button one touch,
after that it may or may not do a normal shutdown.

to make sure, using the config file is given in the unpacked path.

Finally,
fo my test,
bfq is faster and more smoother.

 

[oneghost]

Hey man, thanks a lot for the patches. Can I ask - are these taken from T2SDE svn, or some other source?

 

[Sombat Tragoongoson]

Hello,
testing only on my Gentoo, fine.

 

[oneghost]

Thanks @Sombat Tragoongoson, these are gold

 

[Sombat Tragoongoson]

Hello,

still be fine.

 

[Sombat Tragoongoson]

Hi,

frankly, don't want to update this patch. you can do by yourself form 5.15.

there is other issue to tell.
by my purpose, i don't use VRAM for swapping device.

refer to RE: linux patch update and issue by NRoach44 (15/07/2025)
the issue problem are probably the configuration and ps3fb patch applied.

you have two choices.
1) ps3vram be not used and ps3fb patch applied.
for this configuration, ps3fb patch allocate the whole usable vram size(254M).
so then, the pas3ram cannot be used because vram no space left.
furthermore; you can use my Opengl library kit and xf86-video-ps3.

2) ps3vram be used and ps3fb patch be not applied.
the default the original driver allocate 9M VRAM as kernel configuration.
so then, the ps3vram has 245M available to use for swapping device.
surely, my Opengl library kit and xf86-video-ps3 are not working any more.

I don't yet test the config 2. but the first one is confirmed OK.

so let you choosing.

hello,

sorry for my mistaken,

all of my kernel configs are checked
to CONFIG_PS3_DYNAMIC_DMA.
just first using USB storage,
so then the bug on is arising at
"arch/powerpc/platforms/ps3/system-bus.c".
so please check it off in platform support/PS3 platform options.

 

[Sombat Tragoongoson]

Hello,

linux-6.18, "mount_nodev" is removed from "fs/supper.c".
spuisofs.c and spuldrfs.c both need it.
using the copy form the previous one.
if having another solution, please suggestion.

have fun.