Hi,

Following up on my v2 patch adding vendor-id and device-id properties to QEMU's 
NVMe emulation.

I've cataloged the real-world NVMe quirks from the Linux kernel that this patch 
would help simulate. I have tested these quirks with this patch:

PCI-based Quirks (Vendor ID + Device ID):
  1)  Default Red Hat (0x1b36:0x0010) - BOGUS_NID
  2)  Intel QEMU (0x8086:0x5845) - IDENTIFY_CNS, DISABLE_WRITE_ZEROES, BOGUS_NID
  3)  Intel 750 (0x8086:0x0953) - STRIPE_SIZE, DEALLOCATE_ZEROES
  4)  Intel 600P (0x8086:0xf1a5) - NO_DEEPEST_PS, MEDIUM_PRIO_SQ, 
NO_TEMP_THRESH_CHANGE, DISABLE_WRITE_ZEROES
  5)  Sandisk SN530 (0x15b7:0x5008) - BROKEN_MSI
  6)  Sandisk SN550 (0x15b7:0x5009) - BROKEN_MSI, NO_DEEPEST_PS
  7)  Samsung PM1725 (0x144d:0xa821) - DELAY_BEFORE_CHK_RDY
  8)  Samsung PM1725a (0x144d:0xa822) - DELAY_BEFORE_CHK_RDY, 
DISABLE_WRITE_ZEROES, IGNORE_DEV_SUBNQN
  9)  Phison E12 (0x1987:0x5012) - BOGUS_NID
  10) Phison E19 (0x1987:0x5019) - DISABLE_WRITE_ZEROES
  11) Silicon Motion (0x126f:0x2262) - NO_DEEPEST_PS, BOGUS_NID

Core Quirks (Vendor ID + Model/Firmware):
  12) Toshiba (0x1179 + model) - NO_APST
  13) Samsung X5 (0x144d + model) - DELAY_BEFORE_CHK_RDY, NO_DEEPEST_PS, 
IGNORE_DEV_SUBNQN
  14) Kioxia (0x1e0f + model) - NO_APST
  15) LiteON (0x14a4 + firmware) - SIMPLE_SUSPEND

I'd appreciate any review on the patch.

Thanks,
Utkarsh


Reply via email to