diff mbox series

[v2] nvme: Use only 32-bit accesses in nvme_writeq/nvme_readq

Message ID 27b1d2116c1d37a261bfd6e12f82f84e3e4c079d.1609330566.git.stefan@agner.ch
State Accepted
Commit 71fd11b9013eda5b618737f91e2b19daabb775a3
Delegated to: Tom Rini
Headers show
Series [v2] nvme: Use only 32-bit accesses in nvme_writeq/nvme_readq | expand

Commit Message

Stefan Agner Dec. 30, 2020, 12:16 p.m. UTC
There might be hardware configurations where 64-bit data accesses
to NVMe registers are not supported properly.  This patch removes
the readq/writeq so always two 32-bit accesses are used to read/write
64-bit NVMe registers, similarly as it is done in Linux kernel.

This patch fixes operation of NVMe devices on RPi4 Broadcom BCM2711 SoC
based board, where the PCIe Root Complex, which is attached to the
system through the SCB bridge.

Even though the architecture is 64-bit the PCIe BAR is 32-bit and likely
the 64-bit wide register accesses initiated by the CPU are not properly
translated to a sequence of 32-bit PCIe accesses.
nvme_readq(), for example, always returns same value in upper and lower
32-bits, e.g. 0x3c033fff3c033fff which lead to NVMe devices to fail
probing.

This fix is analogous to commit 8e2ab05000ab ("usb: xhci: Use only
32-bit accesses in xhci_writeq/xhci_readq").

Cc: Sylwester Nawrocki <s.nawrocki@samsung.com>
Cc: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Cc: Matthias Brugger <mbrugger@suse.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Stefan Agner <stefan@agner.ch>
---

Changes in v2:
- Fix subject and message to reflect NVMe change

 drivers/nvme/nvme.h | 8 --------
 1 file changed, 8 deletions(-)

Comments

Tom Rini Jan. 5, 2021, 12:54 p.m. UTC | #1
On Wed, Dec 30, 2020 at 01:16:36PM +0100, Stefan Agner wrote:

> There might be hardware configurations where 64-bit data accesses
> to NVMe registers are not supported properly.  This patch removes
> the readq/writeq so always two 32-bit accesses are used to read/write
> 64-bit NVMe registers, similarly as it is done in Linux kernel.
> 
> This patch fixes operation of NVMe devices on RPi4 Broadcom BCM2711 SoC
> based board, where the PCIe Root Complex, which is attached to the
> system through the SCB bridge.
> 
> Even though the architecture is 64-bit the PCIe BAR is 32-bit and likely
> the 64-bit wide register accesses initiated by the CPU are not properly
> translated to a sequence of 32-bit PCIe accesses.
> nvme_readq(), for example, always returns same value in upper and lower
> 32-bits, e.g. 0x3c033fff3c033fff which lead to NVMe devices to fail
> probing.
> 
> This fix is analogous to commit 8e2ab05000ab ("usb: xhci: Use only
> 32-bit accesses in xhci_writeq/xhci_readq").
> 
> Cc: Sylwester Nawrocki <s.nawrocki@samsung.com>
> Cc: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
> Cc: Matthias Brugger <mbrugger@suse.com>
> Reviewed-by: Stefan Roese <sr@denx.de>
> Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
> Signed-off-by: Stefan Agner <stefan@agner.ch>

Applied to u-boot/master, thanks!
diff mbox series

Patch

diff --git a/drivers/nvme/nvme.h b/drivers/nvme/nvme.h
index 0e8cb221a7..aa4b3bac67 100644
--- a/drivers/nvme/nvme.h
+++ b/drivers/nvme/nvme.h
@@ -535,28 +535,20 @@  struct nvme_completion {
  */
 static inline u64 nvme_readq(__le64 volatile *regs)
 {
-#if BITS_PER_LONG == 64
-	return readq(regs);
-#else
 	__u32 *ptr = (__u32 *)regs;
 	u64 val_lo = readl(ptr);
 	u64 val_hi = readl(ptr + 1);
 
 	return val_lo + (val_hi << 32);
-#endif
 }
 
 static inline void nvme_writeq(const u64 val, __le64 volatile *regs)
 {
-#if BITS_PER_LONG == 64
-	writeq(val, regs);
-#else
 	__u32 *ptr = (__u32 *)regs;
 	u32 val_lo = lower_32_bits(val);
 	u32 val_hi = upper_32_bits(val);
 	writel(val_lo, ptr);
 	writel(val_hi, ptr + 1);
-#endif
 }
 
 struct nvme_bar {