)]}'
{
  "commit": "26ca5f34841f0a86d0883664a1c13be89a01d351",
  "tree": "6c77d4a87bfdfbfb2bc2c7c651b3a6a262da0367",
  "parents": [
    "e9ce64dce975a14700e835003847415e29918474"
  ],
  "author": {
    "name": "Michael Roth",
    "email": "michael.roth@amd.com",
    "time": "Wed Oct 21 14:15:17 2026 -0500"
  },
  "committer": {
    "name": "Herbert Xu",
    "email": "herbert@gondor.apana.org.au",
    "time": "Wed Nov 04 13:46:50 2026 +0800"
  },
  "message": "crypto: ccp - Make usage of sev_cmd_mutex more granular\n\nThe current code makes very coarse usage of the sev_cmd_mutex; for SEV\nfirmware ioctls, it is essentially taken upon entry, and ends up\nwrapping a lot of potentially time-consuming prep work that may need to\nbe done before issuing the actual firmware call, which is the main thing\nthe mutex is meant to protect.\n\nAdditionally, some of that prep work involves\ncopy_from_user()/copy_to_user() calls which can generate page faults and\nunecessarily lead to potential deadlock scenarios. No known case exists\nin current upstream, but with the upcoming KVM/guest_memfd support for\nin-place conversion of the same PFN between shared/private state, a\nspecial fault handler is needed that could theoretically deadlock if a\nmalicious user tries to use guest_memfd pages as the source/target of\npreviously mentioned copy_from_user()/copy_to_user() scenarios.\n\nThere\u0027s no reason to hold the mutex across these problematic scenarios\nhowever. Push sev_cmd_mutex down from the top-level sev_ioctl() into\neach individual command handler, acquiring the mutex mainly around the\nplatform state transitions and firmware command execution.\n\nSuggested-by: Sean Christopherson \u003cseanjc@google.com\u003e\nCo-developed-by: Ackerley Tng \u003cackerleytng@google.com\u003e\nSigned-off-by: Ackerley Tng \u003cackerleytng@google.com\u003e\nSigned-off-by: Michael Roth \u003cmichael.roth@amd.com\u003e\nTested-by: Sean Christopherson \u003cseanjc@google.com\u003e\nReviewed-by: Tom Lendacky \u003cthomas.lendacky@amd.com\u003e\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "5c996ab63895946d753b83a14416ccfda03789e6",
      "old_mode": 33188,
      "old_path": "drivers/crypto/ccp/sev-dev.c",
      "new_id": "b980dfc7f3a33b79b51abcce4fc47892bb1beb32",
      "new_mode": 33188,
      "new_path": "drivers/crypto/ccp/sev-dev.c"
    }
  ]
}
