On Sun, Dec 31, 2023 at 10:33:19AM +0100, Salvatore Bonaccorso wrote:
Hi,
On Sat, Dec 30, 2023 at 11:16:41PM +0000, Jiaxun Yang wrote:
在2023年12月30日十二月 下午7:38，Salvatore Bonaccorso写道：
Hi,
On Mon, Dec 11, 2023 at 07:22:24PM +0100, Greg Kroah-Hartman wrote:
5.10-stable review patch. If anyone has any objections, please let me know.
From: Jiaxun Yang jiaxun.yang@flygoat.com
commit edc0378eee00200a5bedf1bb9f00ad390e0d1bd4 upstream.
There are some Loongson64 systems come with broken coherent DMA support, firmware will set a bit in boot_param and pass nocoherentio in cmdline.
However nonconherent support was missed out when spin off Loongson-2EF form Loongson64, and that boot_param change never made itself into upstream.
Support DMA noncoherent properly to get those systems working.
Cc: stable@vger.kernel.org Fixes: 71e2f4dd5a65 ("MIPS: Fork loongson2ef from loongson64") Signed-off-by: Jiaxun Yang jiaxun.yang@flygoat.com Signed-off-by: Thomas Bogendoerfer tsbogend@alpha.franken.de Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
arch/mips/Kconfig | 2 ++ arch/mips/include/asm/mach-loongson64/boot_param.h | 3 ++- arch/mips/loongson64/env.c | 10 +++++++++- 3 files changed, 13 insertions(+), 2 deletions(-)
--- a/arch/mips/Kconfig +++ b/arch/mips/Kconfig @@ -468,6 +468,7 @@ config MACH_LOONGSON2EF config MACH_LOONGSON64 bool "Loongson 64-bit family of machines"
- select ARCH_DMA_DEFAULT_COHERENT select ARCH_SPARSEMEM_ENABLE select ARCH_MIGHT_HAVE_PC_PARPORT select ARCH_MIGHT_HAVE_PC_SERIO
@@ -1379,6 +1380,7 @@ config CPU_LOONGSON64 select CPU_SUPPORTS_MSA select CPU_DIEI_BROKEN if !LOONGSON3_ENHANCEMENT select CPU_MIPSR2_IRQ_VI
- select DMA_NONCOHERENT select WEAK_ORDERING select WEAK_REORDERING_BEYOND_LLSC select MIPS_ASID_BITS_VARIABLE
--- a/arch/mips/include/asm/mach-loongson64/boot_param.h +++ b/arch/mips/include/asm/mach-loongson64/boot_param.h @@ -117,7 +117,8 @@ struct irq_source_routing_table { u64 pci_io_start_addr; u64 pci_io_end_addr; u64 pci_config_addr;
- u32 dma_mask_bits;
- u16 dma_mask_bits;
- u16 dma_noncoherent;
} __packed; struct interface_info { --- a/arch/mips/loongson64/env.c +++ b/arch/mips/loongson64/env.c @@ -13,6 +13,8 @@
- Copyright (C) 2009 Lemote Inc.
- Author: Wu Zhangjin, wuzhangjin@gmail.com
*/
+#include <linux/dma-map-ops.h> #include <linux/export.h> #include <linux/pci_ids.h> #include <asm/bootinfo.h> @@ -131,8 +133,14 @@ void __init prom_init_env(void) loongson_sysconf.pci_io_base = eirq_source->pci_io_start_addr; loongson_sysconf.dma_mask_bits = eirq_source->dma_mask_bits; if (loongson_sysconf.dma_mask_bits < 32 ||
loongson_sysconf.dma_mask_bits > 64)
loongson_sysconf.dma_mask_bits > 64) {dma_default_coherent = true;- } else {
dma_default_coherent = !eirq_source->dma_noncoherent;- }
- pr_info("Firmware: Coherent DMA: %s\n", dma_default_coherent ? "on" : "off");
loongson_sysconf.restart_addr = boot_p->reset_system.ResetWarm; loongson_sysconf.poweroff_addr = boot_p->reset_system.Shutdown;
While preparing an update for Debian and finally building for all supported architecture, the builds for mipsel and mips64el were failing with:
/<<PKGBUILDDIR>>/arch/mips/loongson64/env.c: In function 'prom_init_env': /<<PKGBUILDDIR>>/arch/mips/loongson64/env.c:138:3: error: 'dma_default_coherent' undeclared (first use in this function); did you mean 'dma_free_coherent'? 138 | dma_default_coherent = true; | ^~~~~~~~~~~~~~~~~~~~ | dma_free_coherent /<<PKGBUILDDIR>>/arch/mips/loongson64/env.c:138:3: note: each undeclared identifier is reported only once for each function it appears in make[6]: *** [/<<PKGBUILDDIR>>/scripts/Makefile.build:291: arch/mips/loongson64/env.o] Error 1 make[6]: *** Waiting for unfinished jobs....
Is here a prerequisite missing for the commit?
Backporting though 6d4e9a8efe3d ("driver core: lift dma_default_coherent into common code") which is from 5.12-rc1 though seems too intrusive, correct? Would the alternative be to just revert the 3ee7e2faef87 ("MIPS: Loongson64: Enable DMA noncoherent support") commit which landed in 5.10.204?
Hi,
Sorry for not spotting the issue in the first place.
It actually requires more dependencies such as c00a60d6f4a1 ("of: address: always use dma_default_coherent for default coherency").
I'll try to produce a backport series but better to drop this patch from stable for now.
Thanks a lot for your quick reply!
So I guess it only needs a revert for now from 5.10.y correct?
Greg, possible to already queue up the revert for 5.10.y for the next upload?
Now reverted, thanks.
greg k-h