Re: [PATCH 0/7] x86/HVM: drop stdvga caching mode

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 10/09/2024 3:38 pm, Jan Beulich wrote:
> It's been unused for nearly 9 years. By the end of the series stdvga.c's
> sole purpose will be to make sure VRAM writes use the bufio ioreq path.
>
> 1: drop stdvga's "cache" struct member
> 2: drop stdvga's "stdvga" struct member
> 3: remove unused MMIO handling code
> 4: drop stdvga's "gr[]" struct member
> 5: drop stdvga's "sr[]" struct member
> 6: drop stdvga's "{g,s}r_index" struct members
> 7: drop stdvga's "vram_page[]" struct member
>
> Of course all of these could be folded into a single patch, but I think
> going piecemeal makes more clear that the individual parts are indeed
> unused.

I far prefer a series split like this, than a single patch with 7 moving
parts.

Also, this is certainly in the right direction.

$ ../scripts/bloat-o-meter -c xen-syms-before xen-syms-after
add/remove: 0/6 grow/shrink: 0/7 up/down: 0/-2640 (-2640)
Function                                     old     new   delta
hvm_domain_destroy                           250     242      -8
hvm_load                                     548     532     -16
hvm_domain_initialise                       1050    1034     -16
stdvga_deinit                                 67       -     -67
stdvga_mem_accept                            236     166     -70
stdvga_mem_offset.isra                       103       -    -103
stdvga_intercept_pio                         158       -    -158
stdvga_init                                  223      52    -171
stdvga_outb.isra                             240       -    -240
stdvga_mem_write                             430     105    -325
stdvga_mem_readb                             396       -    -396
stdvga_mem_read                              423      17    -406
stdvga_mem_writeb                            664       -    -664
add/remove: 0/3 grow/shrink: 0/0 up/down: 0/-81 (-81)
Data                                         old     new   delta
sr_mask                                        8       -      -8
gr_mask                                        9       -      -9
mask16                                        64       -     -64

Interestingly `nm` identifies these as 'd' (initialised data) rather
than 'n' or 'r' (read-only data), but they are between _{s,e}rodata.

~Andrew