On Thu, Jan 30, 2025 at 08:40:15PM -0800, Theodore Ts'o wrote: > On Thu, Jan 30, 2025 at 03:39:20PM -0500, Martin K. Petersen wrote: > > It already works that way. If a device advertises being > > integrity-capable, the block layer will automatically generate > > protection information on write and verify received protection > > information on read. Leveraging hardware-accelerated CRC calculation if > > the CPU is capable (PCLMULQDQ, etc.). > > So I'm confused. If that's the case, why do we need Kanchan Joshi's > patch to set some magic bio flag and adding a mount option to btrfs? Well, as stated in reply to the series there really isn't. The case with a buffer duplicates the existing auto-PI in the block layer, and the no-buffer cases reduces the protection envelope compared to using the auto PI.
