git: fea3cf1d6da0 - main - nvme: Fix alignment on nvme structures

[Warner Losh]

The branch main has been updated by imp:

URL: https://cgit.FreeBSD.org/src/commit/?id=fea3cf1d6da0acf40bc1d3dadeeea7eeccbc10dd

commit fea3cf1d6da0acf40bc1d3dadeeea7eeccbc10dd
Author:     Warner Losh <imp at FreeBSD.org>
AuthorDate: 2021-07-02 21:58:19 +0000
Commit:     Warner Losh <imp at FreeBSD.org>
CommitDate: 2021-07-02 22:05:19 +0000

    nvme: Fix alignment on nvme structures
    
    Remove __packed from nvme_command, nvme_completion and
    nvme_dsm_trim. Add super-alignment to nvme_completion since it's always
    at least that aligned in hardware (and in our existing uses of it
    embedded in structures). It generates better code in
    nvme_qpair_process_completions on riscv64 because otherwise the ABI
    assumes a 4-byte alignment, and the same on all other platforms.
    
    Reviewed by:            jrtc27@, mav@, chuck@
    Sponsored by:           Netflix
    Differential Revision:  https://reviews.freebsd.org/D31001
---
 sys/dev/nvme/nvme.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/sys/dev/nvme/nvme.h b/sys/dev/nvme/nvme.h
index d63a074f321d..7d153c61165c 100644
--- a/sys/dev/nvme/nvme.h
+++ b/sys/dev/nvme/nvme.h
@@ -578,8 +578,8 @@ struct nvme_registers {
 	struct {
 	    uint32_t	sq_tdbl; /* submission queue tail doorbell */
 	    uint32_t	cq_hdbl; /* completion queue head doorbell */
-	} doorbell[1] __packed;
-} __packed;
+	} doorbell[1];
+};
 
 _Static_assert(sizeof(struct nvme_registers) == 0x1008, "bad size for nvme_registers");
 
@@ -612,7 +612,7 @@ struct nvme_command {
 	uint32_t cdw13;		/* command-specific */
 	uint32_t cdw14;		/* command-specific */
 	uint32_t cdw15;		/* command-specific */
-} __packed;
+};
 
 _Static_assert(sizeof(struct nvme_command) == 16 * 4, "bad size for nvme_command");
 
@@ -630,7 +630,7 @@ struct nvme_completion {
 	/* dword 3 */
 	uint16_t		cid;	/* command identifier */
 	uint16_t		status;
-} __packed;
+} __aligned(8);	/* riscv: nvme_qpair_process_completions has better code gen */
 
 _Static_assert(sizeof(struct nvme_completion) == 4 * 4, "bad size for nvme_completion");
 
@@ -638,7 +638,7 @@ struct nvme_dsm_range {
 	uint32_t attributes;
 	uint32_t length;
 	uint64_t starting_lba;
-} __packed;
+};
 
 /* Largest DSM Trim that can be done */
 #define NVME_MAX_DSM_TRIM		4096