[-v3,00/11] SEV: Cleanup sev-guest a bit and add throttling

  From: "Borislav Petkov (AMD)" <bp@alien8.de>

From: "Borislav Petkov (AMD)" <bp@alien8.de>

Hi,

ok, here's v3, rebased ontop of -rc1 and hopefully with all the stuff
addressed. There's always room for improvement but that should come
later and ontop.

Thx.

Changelog:
----------

v2:

ok, here's v2 with the pending stuff fixed.

Thx.

v1:
---
so I've been looking at Dionna's patches adding the SEV guest throttling
and that request issuing spaghetti was getting on my nerves. And it
would've become even worse with more stuff piling ontop so here's
a first round of cleanups before adding more stuff and making it an
unmaintainable mess.

The final result is a lot easier to read with proper separation of
functionality between functions. I want to get rid of more input/output
params being passed back'n'forth and use a struct instead and Nikunj's
patches have stuff which goes in that direction but first things first.

After the cleanup, the new stuff being added is a lot less code and
almost trivial. But I've been staring at this for a while now so it
could be only me who thinks it is trivial now. But we'll see.

Initial smoke testing seems to work ok but I might've introduced some
funky bugs, ofc.

Comments and suggestions are appreciated, as always.

Thanks and thanks, Tom, for the help!


Borislav Petkov (AMD) (7):
  virt/coco/sev-guest: Check SEV_SNP attribute at probe time
  virt/coco/sev-guest: Simplify extended guest request handling
  virt/coco/sev-guest: Remove the disable_vmpck label in
    handle_guest_request()
  virt/coco/sev-guest: Carve out the request issuing logic into a helper
  virt/coco/sev-guest: Do some code style cleanups
  virt/coco/sev-guest: Convert the sw_exit_info_2 checking to a
    switch-case
  crypto: ccp: Get rid of __sev_platform_init_locked()'s local function
    pointer

Dionna Glaze (3):
  virt/coco/sev-guest: Add throttling awareness
  virt/coco/sev-guest: Double-buffer messages
  x86/sev: Change snp_guest_issue_request()'s fw_err argument

Peter Gonda (1):
  crypto: ccp - Name -1 return value as SEV_RET_NO_FW_CALL

 Documentation/virt/coco/sev-guest.rst   |  20 ++-
 arch/x86/include/asm/sev-common.h       |   3 -
 arch/x86/include/asm/sev.h              |  10 +-
 arch/x86/kernel/sev.c                   |  33 ++--
 drivers/crypto/ccp/sev-dev.c            |  22 ++-
 drivers/virt/coco/sev-guest/sev-guest.c | 190 +++++++++++++++---------
 include/uapi/linux/psp-sev.h            |   7 +
 include/uapi/linux/sev-guest.h          |  18 ++-
 8 files changed, 199 insertions(+), 104 deletions(-)
  

On 3/7/23 13:24, Borislav Petkov wrote:
> From: "Borislav Petkov (AMD)" <bp@alien8.de>
> 
> From: "Borislav Petkov (AMD)" <bp@alien8.de>
> 
> Hi,
> 
> ok, here's v3, rebased ontop of -rc1 and hopefully with all the stuff
> addressed. There's always room for improvement but that should come
> later and ontop.
> 
> Thx.

For the series:
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>

One thing I noticed is that patch #3 fixes an issue where the fw_err field 
returns uninitialized stack data when success (0) is returned by the 
ioctl(). There's no Fixes: tag on that commit. But I guess the later 
Fixes: tag on patch #9 would cause #3 to be pulled, too, right?

Thanks,
Tom

> 
> Changelog:
> ----------
> 
> v2:
> 
> ok, here's v2 with the pending stuff fixed.
> 
> Thx.
> 
> v1:
> ---
> so I've been looking at Dionna's patches adding the SEV guest throttling
> and that request issuing spaghetti was getting on my nerves. And it
> would've become even worse with more stuff piling ontop so here's
> a first round of cleanups before adding more stuff and making it an
> unmaintainable mess.
> 
> The final result is a lot easier to read with proper separation of
> functionality between functions. I want to get rid of more input/output
> params being passed back'n'forth and use a struct instead and Nikunj's
> patches have stuff which goes in that direction but first things first.
> 
> After the cleanup, the new stuff being added is a lot less code and
> almost trivial. But I've been staring at this for a while now so it
> could be only me who thinks it is trivial now. But we'll see.
> 
> Initial smoke testing seems to work ok but I might've introduced some
> funky bugs, ofc.
> 
> Comments and suggestions are appreciated, as always.
> 
> Thanks and thanks, Tom, for the help!
> 
> 
> Borislav Petkov (AMD) (7):
>    virt/coco/sev-guest: Check SEV_SNP attribute at probe time
>    virt/coco/sev-guest: Simplify extended guest request handling
>    virt/coco/sev-guest: Remove the disable_vmpck label in
>      handle_guest_request()
>    virt/coco/sev-guest: Carve out the request issuing logic into a helper
>    virt/coco/sev-guest: Do some code style cleanups
>    virt/coco/sev-guest: Convert the sw_exit_info_2 checking to a
>      switch-case
>    crypto: ccp: Get rid of __sev_platform_init_locked()'s local function
>      pointer
> 
> Dionna Glaze (3):
>    virt/coco/sev-guest: Add throttling awareness
>    virt/coco/sev-guest: Double-buffer messages
>    x86/sev: Change snp_guest_issue_request()'s fw_err argument
> 
> Peter Gonda (1):
>    crypto: ccp - Name -1 return value as SEV_RET_NO_FW_CALL
> 
>   Documentation/virt/coco/sev-guest.rst   |  20 ++-
>   arch/x86/include/asm/sev-common.h       |   3 -
>   arch/x86/include/asm/sev.h              |  10 +-
>   arch/x86/kernel/sev.c                   |  33 ++--
>   drivers/crypto/ccp/sev-dev.c            |  22 ++-
>   drivers/virt/coco/sev-guest/sev-guest.c | 190 +++++++++++++++---------
>   include/uapi/linux/psp-sev.h            |   7 +
>   include/uapi/linux/sev-guest.h          |  18 ++-
>   8 files changed, 199 insertions(+), 104 deletions(-)
>
  

On Wed, Mar 08, 2023 at 02:33:58PM -0600, Tom Lendacky wrote:
> One thing I noticed is that patch #3 fixes an issue where the fw_err field
> returns uninitialized stack data when success (0) is returned by the
> ioctl(). There's no Fixes: tag on that commit. But I guess the later Fixes:
> tag on patch #9 would cause #3 to be pulled, too, right?

Yeah, that's the plan...