Re: Installing 11.0_BETA from cd-rom

[Dennis Clarke <dclarke%blastwave.org@localhost>]

On 11/2/25 05:48, Havard Eidnes wrote:
> Hi,
>
> so I had for a long file a Sun Netra T1 installed at a remote
> installation, and then it died. I've now finally diagnosed it,
> and the boot disk was dead-dead, as in "did not show up in
> 'probe-scsi' output from the boot rom"-dead.  So time to fish out
> a spare disk drive, and an occasion to test the installation of
> 11.0_BETA, so that's what I did.

I too have had to deal with some old Sun hardware that was long past
growing long in the tooth and onwards to "please recycle me" stage.
So I snagged one of the new ORACLE S7-2 servers that have fallen out
onto the world most likely due to contract support terms being expired.
Thus far my tests reveal it to be an excellent SPARC server with some
strange processor quirks. The floating point unit is somewhere inside
each of the cores of which we get eight per socket. However there seems
to be eight integer cores per FPU core. I would need to see the actual
sun4v Niagara7 ( or Sonoma in the assembly comments[1] ) to get a feel
for what the heck is going on in there.

<snip>
...
> Anyway, the installation completed, dmesg at
>
>    https://dmesgd.nycbug.org/dmesgd?do=view&id=8682

I am looking forwards to trying the recent NetBSD 11.0 on the S7 machine
and am just curious if SMP is possible or are we limited to a single
core? Here I do not know what a single core really is within the S7
because it arrives in groups of eight cores at a time. None of which
really makes any sense to me but that is quirky sun4v stuff :

neptune# uname -a
SunOS neptune 5.11 11.4.81.193.1 sun4v sparc sun4v non-virtualized
neptune# psrinfo -pv
The physical processor has 2 cores and 16 virtual processors (0-15)
  The core has 8 virtual processors (0-7)
  The core has 8 virtual processors (8-15)
    SPARC-S7 (chipid 0, clock 4267 MHz)
neptune#
neptune# ldm list
NAME      STATE   FLAGS   CONS  VCPU  MEMORY   UTIL  NORM  UPTIME
primary   active  -n-cv-  UART  16    64G      0.1%  0.1%  13d 22h 16m
s7gentoo  active  -n----  5000  112   190G     0.9%  0.9%  1d 16h 13m
neptune#

So the primary LDOM for Solaris 11.4 is restricted it down to 16 cores
but really it is just two. Weird. The 64G of memory is needed in order
to handle the ZFS stuff.

So for NetBSD the s7gentoo can be reduced down a tad and then what?


--
--
Dennis Clarke
RISC-V/SPARC/PPC/ARM/CISC
UNIX and Linux spoken

[1] seen in a patch for Linux arch/sparc/include/asm/spitfire.h

000043  #define SUN4V_CHIP_INVALID      0x00
000044  #define SUN4V_CHIP_NIAGARA1     0x01
000045  #define SUN4V_CHIP_NIAGARA2     0x02
000046  #define SUN4V_CHIP_NIAGARA3     0x03
000047  #define SUN4V_CHIP_NIAGARA4     0x04
000048  #define SUN4V_CHIP_NIAGARA5     0x05
000049  #define SUN4V_CHIP_SPARC_M6     0x06
000050  #define SUN4V_CHIP_SPARC_M7     0x07
000051  #define SUN4V_CHIP_SPARC_M8     0x08
000052  #define SUN4V_CHIP_SPARC_S7     0x7a
000053  #define SUN4V_CHIP_SPARC64X     0x8a
000054  #define SUN4V_CHIP_SPARC_SN     0x8b
000055  #define SUN4V_CHIP_UNKNOWN      0xff
000056
000057  /*
000058   * The following CPU_ID_xxx constants are used
000059   * to identify the CPU type in the setup phase
000060   * (see head_64.S)
000061   */
000062  #define CPU_ID_NIAGARA1         ('1')
000063  #define CPU_ID_NIAGARA2         ('2')
000064  #define CPU_ID_NIAGARA3         ('3')
000065  #define CPU_ID_NIAGARA4         ('4')
000066  #define CPU_ID_NIAGARA5         ('5')
000067  #define CPU_ID_M6               ('6')
000068  #define CPU_ID_M7               ('7')
000069  #define CPU_ID_M8               ('8')
000070  /* The SONOMA seems to be the S7 as seen in
000071   * arch/sparc/kernel/cpu.c around line 517
000072   */
000073  #define CPU_ID_SONOMA1          ('N')
000074

Then in various places we see the S7 referred to as SONOMA.