Kernel
Threads by month
- ----- 2025 -----
- June
- May
- April
- March
- February
- January
- ----- 2024 -----
- December
- November
- October
- September
- August
- July
- June
- May
- April
- March
- February
- January
- ----- 2023 -----
- December
- November
- October
- September
- August
- July
- June
- May
- April
- March
- February
- January
- ----- 2022 -----
- December
- November
- October
- September
- August
- July
- June
- May
- April
- March
- February
- January
- ----- 2021 -----
- December
- November
- October
- September
- August
- July
- June
- May
- April
- March
- February
- January
- ----- 2020 -----
- December
- November
- October
- September
- August
- July
- June
- May
- April
- March
- February
- January
- ----- 2019 -----
- December
- 18444 discussions
[PATCH OLK-6.6] media: venus: hfi_parser: refactor hfi packet parsing logic
by Xiongfeng Wang 30 May '25
by Xiongfeng Wang 30 May '25
30 May '25
From: Vikash Garodia <quic_vgarodia(a)quicinc.com>
stable inclusion
from stable-v6.6.88
commit 05b07e52a0d08239147ba3460045855f4fb398de
category: bugfix
bugzilla: https://gitee.com/src-openeuler/kernel/issues/IC5BJ2
CVE: CVE-2025-23156
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id…
--------------------------------
commit 9edaaa8e3e15aab1ca413ab50556de1975bcb329 upstream.
words_count denotes the number of words in total payload, while data
points to payload of various property within it. When words_count
reaches last word, data can access memory beyond the total payload. This
can lead to OOB access. With this patch, the utility api for handling
individual properties now returns the size of data consumed. Accordingly
remaining bytes are calculated before parsing the payload, thereby
eliminates the OOB access possibilities.
Cc: stable(a)vger.kernel.org
Fixes: 1a73374a04e5 ("media: venus: hfi_parser: add common capability parser")
Signed-off-by: Vikash Garodia <quic_vgarodia(a)quicinc.com>
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue(a)linaro.org>
Signed-off-by: Hans Verkuil <hverkuil(a)xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org>
Signed-off-by: Wang Hai <wanghai38(a)huawei.com>
Signed-off-by: Xiongfeng Wang <wangxiongfeng2(a)huawei.com>
---
.../media/platform/qcom/venus/hfi_parser.c | 98 ++++++++++++++-----
1 file changed, 72 insertions(+), 26 deletions(-)
diff --git a/drivers/media/platform/qcom/venus/hfi_parser.c b/drivers/media/platform/qcom/venus/hfi_parser.c
index c43839539d4d..aa7819d21f99 100644
--- a/drivers/media/platform/qcom/venus/hfi_parser.c
+++ b/drivers/media/platform/qcom/venus/hfi_parser.c
@@ -62,7 +62,7 @@ fill_buf_mode(struct hfi_plat_caps *cap, const void *data, unsigned int num)
cap->cap_bufs_mode_dynamic = true;
}
-static void
+static int
parse_alloc_mode(struct venus_core *core, u32 codecs, u32 domain, void *data)
{
struct hfi_buffer_alloc_mode_supported *mode = data;
@@ -70,7 +70,7 @@ parse_alloc_mode(struct venus_core *core, u32 codecs, u32 domain, void *data)
u32 *type;
if (num_entries > MAX_ALLOC_MODE_ENTRIES)
- return;
+ return -EINVAL;
type = mode->data;
@@ -82,6 +82,8 @@ parse_alloc_mode(struct venus_core *core, u32 codecs, u32 domain, void *data)
type++;
}
+
+ return sizeof(*mode);
}
static void fill_profile_level(struct hfi_plat_caps *cap, const void *data,
@@ -96,7 +98,7 @@ static void fill_profile_level(struct hfi_plat_caps *cap, const void *data,
cap->num_pl += num;
}
-static void
+static int
parse_profile_level(struct venus_core *core, u32 codecs, u32 domain, void *data)
{
struct hfi_profile_level_supported *pl = data;
@@ -104,12 +106,14 @@ parse_profile_level(struct venus_core *core, u32 codecs, u32 domain, void *data)
struct hfi_profile_level pl_arr[HFI_MAX_PROFILE_COUNT] = {};
if (pl->profile_count > HFI_MAX_PROFILE_COUNT)
- return;
+ return -EINVAL;
memcpy(pl_arr, proflevel, pl->profile_count * sizeof(*proflevel));
for_each_codec(core->caps, ARRAY_SIZE(core->caps), codecs, domain,
fill_profile_level, pl_arr, pl->profile_count);
+
+ return pl->profile_count * sizeof(*proflevel) + sizeof(u32);
}
static void
@@ -124,7 +128,7 @@ fill_caps(struct hfi_plat_caps *cap, const void *data, unsigned int num)
cap->num_caps += num;
}
-static void
+static int
parse_caps(struct venus_core *core, u32 codecs, u32 domain, void *data)
{
struct hfi_capabilities *caps = data;
@@ -133,12 +137,14 @@ parse_caps(struct venus_core *core, u32 codecs, u32 domain, void *data)
struct hfi_capability caps_arr[MAX_CAP_ENTRIES] = {};
if (num_caps > MAX_CAP_ENTRIES)
- return;
+ return -EINVAL;
memcpy(caps_arr, cap, num_caps * sizeof(*cap));
for_each_codec(core->caps, ARRAY_SIZE(core->caps), codecs, domain,
fill_caps, caps_arr, num_caps);
+
+ return sizeof(*caps);
}
static void fill_raw_fmts(struct hfi_plat_caps *cap, const void *fmts,
@@ -153,7 +159,7 @@ static void fill_raw_fmts(struct hfi_plat_caps *cap, const void *fmts,
cap->num_fmts += num_fmts;
}
-static void
+static int
parse_raw_formats(struct venus_core *core, u32 codecs, u32 domain, void *data)
{
struct hfi_uncompressed_format_supported *fmt = data;
@@ -162,7 +168,8 @@ parse_raw_formats(struct venus_core *core, u32 codecs, u32 domain, void *data)
struct raw_formats rawfmts[MAX_FMT_ENTRIES] = {};
u32 entries = fmt->format_entries;
unsigned int i = 0;
- u32 num_planes;
+ u32 num_planes = 0;
+ u32 size;
while (entries) {
num_planes = pinfo->num_planes;
@@ -172,7 +179,7 @@ parse_raw_formats(struct venus_core *core, u32 codecs, u32 domain, void *data)
i++;
if (i >= MAX_FMT_ENTRIES)
- return;
+ return -EINVAL;
if (pinfo->num_planes > MAX_PLANES)
break;
@@ -184,9 +191,13 @@ parse_raw_formats(struct venus_core *core, u32 codecs, u32 domain, void *data)
for_each_codec(core->caps, ARRAY_SIZE(core->caps), codecs, domain,
fill_raw_fmts, rawfmts, i);
+ size = fmt->format_entries * (sizeof(*constr) * num_planes + 2 * sizeof(u32))
+ + 2 * sizeof(u32);
+
+ return size;
}
-static void parse_codecs(struct venus_core *core, void *data)
+static int parse_codecs(struct venus_core *core, void *data)
{
struct hfi_codec_supported *codecs = data;
@@ -198,21 +209,27 @@ static void parse_codecs(struct venus_core *core, void *data)
core->dec_codecs &= ~HFI_VIDEO_CODEC_SPARK;
core->enc_codecs &= ~HFI_VIDEO_CODEC_HEVC;
}
+
+ return sizeof(*codecs);
}
-static void parse_max_sessions(struct venus_core *core, const void *data)
+static int parse_max_sessions(struct venus_core *core, const void *data)
{
const struct hfi_max_sessions_supported *sessions = data;
core->max_sessions_supported = sessions->max_sessions;
+
+ return sizeof(*sessions);
}
-static void parse_codecs_mask(u32 *codecs, u32 *domain, void *data)
+static int parse_codecs_mask(u32 *codecs, u32 *domain, void *data)
{
struct hfi_codec_mask_supported *mask = data;
*codecs = mask->codecs;
*domain = mask->video_domains;
+
+ return sizeof(*mask);
}
static void parser_init(struct venus_inst *inst, u32 *codecs, u32 *domain)
@@ -281,8 +298,9 @@ static int hfi_platform_parser(struct venus_core *core, struct venus_inst *inst)
u32 hfi_parser(struct venus_core *core, struct venus_inst *inst, void *buf,
u32 size)
{
- unsigned int words_count = size >> 2;
- u32 *word = buf, *data, codecs = 0, domain = 0;
+ u32 *words = buf, *payload, codecs = 0, domain = 0;
+ u32 *frame_size = buf + size;
+ u32 rem_bytes = size;
int ret;
ret = hfi_platform_parser(core, inst);
@@ -299,38 +317,66 @@ u32 hfi_parser(struct venus_core *core, struct venus_inst *inst, void *buf,
memset(core->caps, 0, sizeof(core->caps));
}
- while (words_count) {
- data = word + 1;
+ while (words < frame_size) {
+ payload = words + 1;
- switch (*word) {
+ switch (*words) {
case HFI_PROPERTY_PARAM_CODEC_SUPPORTED:
- parse_codecs(core, data);
+ if (rem_bytes <= sizeof(struct hfi_codec_supported))
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ ret = parse_codecs(core, payload);
+ if (ret < 0)
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
init_codecs(core);
break;
case HFI_PROPERTY_PARAM_MAX_SESSIONS_SUPPORTED:
- parse_max_sessions(core, data);
+ if (rem_bytes <= sizeof(struct hfi_max_sessions_supported))
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ ret = parse_max_sessions(core, payload);
break;
case HFI_PROPERTY_PARAM_CODEC_MASK_SUPPORTED:
- parse_codecs_mask(&codecs, &domain, data);
+ if (rem_bytes <= sizeof(struct hfi_codec_mask_supported))
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ ret = parse_codecs_mask(&codecs, &domain, payload);
break;
case HFI_PROPERTY_PARAM_UNCOMPRESSED_FORMAT_SUPPORTED:
- parse_raw_formats(core, codecs, domain, data);
+ if (rem_bytes <= sizeof(struct hfi_uncompressed_format_supported))
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ ret = parse_raw_formats(core, codecs, domain, payload);
break;
case HFI_PROPERTY_PARAM_CAPABILITY_SUPPORTED:
- parse_caps(core, codecs, domain, data);
+ if (rem_bytes <= sizeof(struct hfi_capabilities))
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ ret = parse_caps(core, codecs, domain, payload);
break;
case HFI_PROPERTY_PARAM_PROFILE_LEVEL_SUPPORTED:
- parse_profile_level(core, codecs, domain, data);
+ if (rem_bytes <= sizeof(struct hfi_profile_level_supported))
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ ret = parse_profile_level(core, codecs, domain, payload);
break;
case HFI_PROPERTY_PARAM_BUFFER_ALLOC_MODE_SUPPORTED:
- parse_alloc_mode(core, codecs, domain, data);
+ if (rem_bytes <= sizeof(struct hfi_buffer_alloc_mode_supported))
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ ret = parse_alloc_mode(core, codecs, domain, payload);
break;
default:
+ ret = sizeof(u32);
break;
}
- word++;
- words_count--;
+ if (ret < 0)
+ return HFI_ERR_SYS_INSUFFICIENT_RESOURCES;
+
+ words += ret / sizeof(u32);
+ rem_bytes -= ret;
}
if (!core->max_sessions_supported)
--
2.20.1
2
1
[PATCH OLK-6.6] soc: samsung: exynos-chipid: Add NULL pointer check in exynos_chipid_probe()
by Xiongfeng Wang 30 May '25
by Xiongfeng Wang 30 May '25
30 May '25
From: Chenyuan Yang <chenyuan0y(a)gmail.com>
stable inclusion
from stable-v6.6.88
commit 5f80fd2ff8bfd13e41554741740e0ca8e6445ded
category: bugfix
bugzilla: https://gitee.com/src-openeuler/kernel/issues/IC5BL6
CVE: CVE-2025-23148
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id…
--------------------------------
commit c8222ef6cf29dd7cad21643228f96535cc02b327 upstream.
soc_dev_attr->revision could be NULL, thus,
a pointer check is added to prevent potential NULL pointer dereference.
This is similar to the fix in commit 3027e7b15b02
("ice: Fix some null pointer dereference issues in ice_ptp.c").
This issue is found by our static analysis tool.
Signed-off-by: Chenyuan Yang <chenyuan0y(a)gmail.com>
Link: https://lore.kernel.org/r/20250212213518.69432-1-chenyuan0y@gmail.com
Fixes: 3253b7b7cd44 ("soc: samsung: Add exynos chipid driver support")
Cc: <stable(a)vger.kernel.org>
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski(a)linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org>
Signed-off-by: Wang Hai <wanghai38(a)huawei.com>
Signed-off-by: Xiongfeng Wang <wangxiongfeng2(a)huawei.com>
---
drivers/soc/samsung/exynos-chipid.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/soc/samsung/exynos-chipid.c b/drivers/soc/samsung/exynos-chipid.c
index 7ba45c4aff97..1307f6d61c98 100644
--- a/drivers/soc/samsung/exynos-chipid.c
+++ b/drivers/soc/samsung/exynos-chipid.c
@@ -130,6 +130,8 @@ static int exynos_chipid_probe(struct platform_device *pdev)
soc_dev_attr->revision = devm_kasprintf(&pdev->dev, GFP_KERNEL,
"%x", soc_info.revision);
+ if (!soc_dev_attr->revision)
+ return -ENOMEM;
soc_dev_attr->soc_id = product_id_to_soc_id(soc_info.product_id);
if (!soc_dev_attr->soc_id) {
pr_err("Unknown SoC\n");
--
2.20.1
2
1
30 May '25
mainline inclusion
from mainline-v6.15-rc1
commit 1b0449544c6482179ac84530b61fc192a6527bfd
category: bugfix
bugzilla: https://gitee.com/src-openeuler/kernel/issues/IC6BDN
CVE: CVE-2025-37834
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?…
--------------------------------
Syzkaller reports a bug as follows:
Injecting memory failure for pfn 0x18b00e at process virtual address 0x20ffd000
Memory failure: 0x18b00e: dirty swapcache page still referenced by 2 users
Memory failure: 0x18b00e: recovery action for dirty swapcache page: Failed
page: refcount:2 mapcount:0 mapping:0000000000000000 index:0x20ffd pfn:0x18b00e
memcg:ffff0000dd6d9000
anon flags: 0x5ffffe00482011(locked|dirty|arch_1|swapbacked|hwpoison|node=0|zone=2|lastcpupid=0xfffff)
raw: 005ffffe00482011 dead000000000100 dead000000000122 ffff0000e232a7c9
raw: 0000000000020ffd 0000000000000000 00000002ffffffff ffff0000dd6d9000
page dumped because: VM_BUG_ON_FOLIO(!folio_test_uptodate(folio))
------------[ cut here ]------------
kernel BUG at mm/swap_state.c:184!
Internal error: Oops - BUG: 00000000f2000800 [#1] SMP
Modules linked in:
CPU: 0 PID: 60 Comm: kswapd0 Not tainted 6.6.0-gcb097e7de84e #3
Hardware name: linux,dummy-virt (DT)
pstate: 80400005 (Nzcv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : add_to_swap+0xbc/0x158
lr : add_to_swap+0xbc/0x158
sp : ffff800087f37340
x29: ffff800087f37340 x28: fffffc00052c0380 x27: ffff800087f37780
x26: ffff800087f37490 x25: ffff800087f37c78 x24: ffff800087f377a0
x23: ffff800087f37c50 x22: 0000000000000000 x21: fffffc00052c03b4
x20: 0000000000000000 x19: fffffc00052c0380 x18: 0000000000000000
x17: 296f696c6f662865 x16: 7461646f7470755f x15: 747365745f6f696c
x14: 6f6621284f494c4f x13: 0000000000000001 x12: ffff600036d8b97b
x11: 1fffe00036d8b97a x10: ffff600036d8b97a x9 : dfff800000000000
x8 : 00009fffc9274686 x7 : ffff0001b6c5cbd3 x6 : 0000000000000001
x5 : ffff0000c25896c0 x4 : 0000000000000000 x3 : 0000000000000000
x2 : 0000000000000000 x1 : ffff0000c25896c0 x0 : 0000000000000000
Call trace:
add_to_swap+0xbc/0x158
shrink_folio_list+0x12ac/0x2648
shrink_inactive_list+0x318/0x948
shrink_lruvec+0x450/0x720
shrink_node_memcgs+0x280/0x4a8
shrink_node+0x128/0x978
balance_pgdat+0x4f0/0xb20
kswapd+0x228/0x438
kthread+0x214/0x230
ret_from_fork+0x10/0x20
I can reproduce this issue with the following steps:
1) When a dirty swapcache page is isolated by reclaim process and the
page isn't locked, inject memory failure for the page.
me_swapcache_dirty() clears uptodate flag and tries to delete from lru,
but fails. Reclaim process will put the hwpoisoned page back to lru.
2) The process that maps the hwpoisoned page exits, the page is deleted
the page will never be freed and will be in the lru forever.
3) If we trigger a reclaim again and tries to reclaim the page,
add_to_swap() will trigger VM_BUG_ON_FOLIO due to the uptodate flag is
cleared.
To fix it, skip the hwpoisoned page in shrink_folio_list(). Besides, the
hwpoison folio may not be unmapped by hwpoison_user_mappings() yet, unmap
it in shrink_folio_list(), otherwise the folio will fail to be unmaped by
hwpoison_user_mappings() since the folio isn't in lru list.
Link: https://lkml.kernel.org/r/20250318083939.987651-3-tujinjiang@huawei.com
Signed-off-by: Jinjiang Tu <tujinjiang(a)huawei.com>
Acked-by: Miaohe Lin <linmiaohe(a)huawei.com>
Cc: David Hildenbrand <david(a)redhat.com>
Cc: Kefeng Wang <wangkefeng.wang(a)huawei.com>
Cc: Nanyong Sun <sunnanyong(a)huawei.com>
Cc: Naoya Horiguchi <nao.horiguchi(a)gmail.com>
Cc: <stable(a)vger,kernel.org>
Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org>
Conflicts:
mm/vmscan.c
[Many conflicts, fix it with same idea.]
Signed-off-by: Jinjiang Tu <tujinjiang(a)huawei.com>
---
mm/vmscan.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 40bcde9d56be..df7998e8174f 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1169,6 +1169,13 @@ static unsigned long shrink_page_list(struct list_head *page_list,
if (!trylock_page(page))
goto keep;
+ if (PageHWPoison(page)) {
+ try_to_unmap(page, TTU_IGNORE_MLOCK | TTU_IGNORE_ACCESS);
+ unlock_page(page);
+ put_page(page);
+ continue;
+ }
+
VM_BUG_ON_PAGE(PageActive(page), page);
sc->nr_scanned++;
--
2.43.0
2
1
30 May '25
hulk inclusion
category: feature
bugzilla: https://gitee.com/openeuler/kernel/issues/ICBFCS
CVE: NA
--------------------------------
This introduces a kind of network optimization method named oenetcls. It
can configure the ntuple rule, and bind interrupt to the netdev queue
automatically.
Signed-off-by: Yue Haibing <yuehaibing(a)huawei.com>
Signed-off-by: Wang Liang <wangliang74(a)huawei.com>
Signed-off-by: Liu Jian <liujian56(a)huawei.com>
Signed-off-by: yuelg <yuelg(a)chinaunicom.cn>
---
drivers/hooks/Kconfig | 10 +
drivers/hooks/vendor_hooks.c | 8 +
include/linux/netdevice.h | 2 +-
include/trace/hooks/oenetcls.h | 44 ++
net/Kconfig | 1 +
net/Makefile | 1 +
net/core/dev.c | 21 +
net/ipv4/af_inet.c | 5 +
net/ipv4/tcp.c | 10 +
net/oenetcls/Kconfig | 10 +
net/oenetcls/Makefile | 8 +
net/oenetcls/asmdefs.h | 61 +++
net/oenetcls/memcpy-sve.S | 157 ++++++
net/oenetcls/oenetcls.h | 176 ++++++
net/oenetcls/oenetcls_flow.c | 402 ++++++++++++++
net/oenetcls/oenetcls_l0.c | 193 +++++++
net/oenetcls/oenetcls_main.c | 972 +++++++++++++++++++++++++++++++++
net/oenetcls/oenetcls_ntuple.c | 510 +++++++++++++++++
18 files changed, 2590 insertions(+), 1 deletion(-)
create mode 100644 include/trace/hooks/oenetcls.h
create mode 100644 net/oenetcls/Kconfig
create mode 100644 net/oenetcls/Makefile
create mode 100644 net/oenetcls/asmdefs.h
create mode 100644 net/oenetcls/memcpy-sve.S
create mode 100644 net/oenetcls/oenetcls.h
create mode 100644 net/oenetcls/oenetcls_flow.c
create mode 100644 net/oenetcls/oenetcls_l0.c
create mode 100644 net/oenetcls/oenetcls_main.c
create mode 100644 net/oenetcls/oenetcls_ntuple.c
diff --git a/drivers/hooks/Kconfig b/drivers/hooks/Kconfig
index 6a00168e67ad..90b0f6ea4040 100644
--- a/drivers/hooks/Kconfig
+++ b/drivers/hooks/Kconfig
@@ -20,4 +20,14 @@ config VENDOR_BOND_HOOKS
Allow vendor modules to attach bonding driver hooks defined via
DECLARE_HOOK or DECLARE_RESTRICTED_HOOK.
+config OENETCLS_HOOKS
+ bool "Oenetcls driver Hooks"
+ depends on VENDOR_HOOKS
+ default n
+ help
+ Enable oenetcls vendor hooks
+ Allow vendor modules to attach oenetcls hooks defined via
+ DECLARE_HOOK or DECLARE_RESTRICTED_HOOK.
+ Use OENETCLS && OENETCLS_HOOKS to enable oenetcls feature.
+
endmenu
diff --git a/drivers/hooks/vendor_hooks.c b/drivers/hooks/vendor_hooks.c
index 85bda58159f6..d9b85b57a742 100644
--- a/drivers/hooks/vendor_hooks.c
+++ b/drivers/hooks/vendor_hooks.c
@@ -9,6 +9,7 @@
#define CREATE_TRACE_POINTS
#include <trace/hooks/vendor_hooks.h>
#include <trace/hooks/bonding.h>
+#include <trace/hooks/oenetcls.h>
/*
* Export tracepoints that act as a bare tracehook (ie: have no trace event
@@ -18,3 +19,10 @@
#ifdef CONFIG_VENDOR_BOND_HOOKS
EXPORT_TRACEPOINT_SYMBOL_GPL(vendor_bond_check_dev_link);
#endif
+
+#ifdef CONFIG_OENETCLS_HOOKS
+EXPORT_TRACEPOINT_SYMBOL_GPL(oecls_flow_update);
+EXPORT_TRACEPOINT_SYMBOL_GPL(oecls_set_cpu);
+EXPORT_TRACEPOINT_SYMBOL_GPL(oecls_timeout);
+EXPORT_TRACEPOINT_SYMBOL_GPL(ethtool_cfg_rxcls);
+#endif
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index df61a63a5550..4b7b6fa1dcb1 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -808,7 +808,7 @@ struct netdev_rx_queue {
struct xsk_buff_pool *pool;
#endif
- KABI_RESERVE(1)
+ KABI_USE(1, void *__rcu oecls_ftb)
KABI_RESERVE(2)
KABI_RESERVE(3)
KABI_RESERVE(4)
diff --git a/include/trace/hooks/oenetcls.h b/include/trace/hooks/oenetcls.h
new file mode 100644
index 000000000000..c38545d7a6a2
--- /dev/null
+++ b/include/trace/hooks/oenetcls.h
@@ -0,0 +1,44 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * oenetcls driver Hooks
+ *
+ * Copyright (c) 2025, Huawei Tech. Co., Ltd.
+ */
+
+#ifdef CONFIG_OENETCLS_HOOKS
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM oenetcls
+
+#define TRACE_INCLUDE_PATH trace/hooks
+#if !defined(_TRACE_OENETCLS_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_OENETCLS_H
+#include <linux/tracepoint.h>
+#include <trace/hooks/vendor_hooks.h>
+
+struct sock;
+struct sk_buff;
+struct net_device;
+
+DECLARE_HOOK(oecls_flow_update,
+TP_PROTO(struct sock *sk),
+TP_ARGS(sk));
+
+DECLARE_HOOK(oecls_set_cpu,
+TP_PROTO(struct sk_buff *skb),
+TP_ARGS(skb));
+
+DECLARE_HOOK(oecls_timeout,
+TP_PROTO(struct net_device *dev, u16 rxq_index, u32 flow_id, u16 filter_id, bool *ret),
+TP_ARGS(dev, rxq_index, flow_id, filter_id, ret));
+
+DECLARE_HOOK(ethtool_cfg_rxcls,
+TP_PROTO(struct sock *sk, int is_del),
+TP_ARGS(sk, is_del));
+
+#endif
+/* This part must be outside protection */
+#include <trace/define_trace.h>
+
+#endif
+
diff --git a/net/Kconfig b/net/Kconfig
index 51a934426f9f..071f8ee3b89f 100644
--- a/net/Kconfig
+++ b/net/Kconfig
@@ -67,6 +67,7 @@ source "net/xfrm/Kconfig"
source "net/iucv/Kconfig"
source "net/smc/Kconfig"
source "net/xdp/Kconfig"
+source "net/oenetcls/Kconfig"
config INET
bool "TCP/IP networking"
diff --git a/net/Makefile b/net/Makefile
index 45c03aa92ace..dc637db866e3 100644
--- a/net/Makefile
+++ b/net/Makefile
@@ -87,3 +87,4 @@ obj-$(CONFIG_QRTR) += qrtr/
obj-$(CONFIG_NET_NCSI) += ncsi/
obj-$(CONFIG_XDP_SOCKETS) += xdp/
obj-$(CONFIG_MPTCP) += mptcp/
+obj-$(CONFIG_OENETCLS) += oenetcls/
diff --git a/net/core/dev.c b/net/core/dev.c
index c49f3d306b5c..eda1975ef55b 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -149,6 +149,7 @@
#include <net/net_rship.h>
#include "net-sysfs.h"
+#include <trace/hooks/oenetcls.h>
#define MAX_GRO_SKBS 8
@@ -4484,6 +4485,11 @@ bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
bool expire = true;
unsigned int cpu;
+#if IS_ENABLED(CONFIG_OENETCLS_HOOKS)
+ trace_oecls_timeout(dev, rxq_index, flow_id, filter_id, &expire);
+ if (expire)
+ return true;
+#endif
rcu_read_lock();
flow_table = rcu_dereference(rxqueue->rps_flow_table);
if (flow_table && flow_id <= flow_table->mask) {
@@ -5819,6 +5825,11 @@ static int netif_receive_skb_internal(struct sk_buff *skb)
}
}
#endif
+
+#if IS_ENABLED(CONFIG_OENETCLS_HOOKS)
+ trace_oecls_set_cpu(skb);
+#endif
+
ret = __netif_receive_skb(skb);
rcu_read_unlock();
return ret;
@@ -5853,6 +5864,12 @@ static void netif_receive_skb_list_internal(struct list_head *head)
}
}
#endif
+
+#if IS_ENABLED(CONFIG_OENETCLS_HOOKS)
+ list_for_each_entry_safe(skb, next, head, list)
+ trace_oecls_set_cpu(skb);
+#endif
+
__netif_receive_skb_list(head);
rcu_read_unlock();
}
@@ -9984,6 +10001,10 @@ int __netdev_update_features(struct net_device *dev)
return err < 0 ? 0 : 1;
}
+#if IS_ENABLED(CONFIG_OENETCLS)
+EXPORT_SYMBOL(__netdev_update_features);
+#endif
+
/**
* netdev_update_features - recalculate device features
* @dev: the device to check
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index c6ec22f314c4..b09b1d6575a4 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -119,6 +119,7 @@
#include <net/compat.h>
#include <trace/events/sock.h>
+#include <trace/hooks/oenetcls.h>
/* The inetsw table contains everything that inet_create needs to
* build a new socket.
@@ -232,6 +233,10 @@ int inet_listen(struct socket *sock, int backlog)
if (err)
goto out;
tcp_call_bpf(sk, BPF_SOCK_OPS_TCP_LISTEN_CB, 0, NULL);
+
+#if IS_ENABLED(CONFIG_OENETCLS_HOOKS)
+ trace_ethtool_cfg_rxcls(sk, 0);
+#endif
}
err = 0;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 2a5147a4d1af..f014fad55bbf 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -280,6 +280,7 @@
#include <asm/ioctls.h>
#include <net/busy_poll.h>
#include <net/net_rship.h>
+#include <trace/hooks/oenetcls.h>
DEFINE_PER_CPU(unsigned int, tcp_orphan_count);
EXPORT_PER_CPU_SYMBOL_GPL(tcp_orphan_count);
@@ -2176,6 +2177,10 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
if (unlikely(flags & MSG_ERRQUEUE))
return inet_recv_error(sk, msg, len, addr_len);
+#if IS_ENABLED(CONFIG_OENETCLS_HOOKS)
+ trace_oecls_flow_update(sk);
+#endif
+
if (sk_can_busy_loop(sk) && skb_queue_empty_lockless(&sk->sk_receive_queue) &&
(sk->sk_state == TCP_ESTABLISHED))
sk_busy_loop(sk, nonblock);
@@ -2750,6 +2755,11 @@ void __tcp_close(struct sock *sk, long timeout)
void tcp_close(struct sock *sk, long timeout)
{
lock_sock(sk);
+
+#if IS_ENABLED(CONFIG_OENETCLS_HOOKS)
+ trace_ethtool_cfg_rxcls(sk, 1);
+#endif
+
__tcp_close(sk, timeout);
release_sock(sk);
if (!sk->sk_net_refcnt)
diff --git a/net/oenetcls/Kconfig b/net/oenetcls/Kconfig
new file mode 100644
index 000000000000..128b798d6b7f
--- /dev/null
+++ b/net/oenetcls/Kconfig
@@ -0,0 +1,10 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config OENETCLS
+ tristate "Network classification"
+ depends on HISI_L0 && OENETCLS_HOOKS
+ default n
+ help
+ Allows to configure ntuple rule, and bind interrupt to netdev
+ automatically.
+ Use OENETCLS && OENETCLS_HOOKS to enable oenetcls feature.
+ Use parameter mode to decide running mode.
diff --git a/net/oenetcls/Makefile b/net/oenetcls/Makefile
new file mode 100644
index 000000000000..4d395d36a5fc
--- /dev/null
+++ b/net/oenetcls/Makefile
@@ -0,0 +1,8 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+obj-$(CONFIG_OENETCLS) = oenetcls.o
+oenetcls-y := oenetcls_main.o oenetcls_ntuple.o oenetcls_flow.o oenetcls_l0.o
+ifeq ($(CONFIG_ARM64_SVE),y)
+oenetcls-y += memcpy-sve.o
+endif
+
diff --git a/net/oenetcls/asmdefs.h b/net/oenetcls/asmdefs.h
new file mode 100644
index 000000000000..8138a94c18af
--- /dev/null
+++ b/net/oenetcls/asmdefs.h
@@ -0,0 +1,61 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef _ASMDEFS_H
+#define _ASMDEFS_H
+
+/* Branch Target Identitication support. */
+#define BTI_C hint 34
+#define BTI_J hint 36
+/* Return address signing support (pac-ret). */
+#define PACIASP hint 25; .cfi_window_save
+#define AUTIASP hint 29; .cfi_window_save
+
+/* GNU_PROPERTY_AARCH64_* macros from elf.h. */
+#define FEATURE_1_AND 0xc0000000
+#define FEATURE_1_BTI 1
+#define FEATURE_1_PAC 2
+
+/* Add a NT_GNU_PROPERTY_TYPE_0 note. */
+#define GNU_PROPERTY(type, value) \
+ .section .note.gnu.property, "a"; \
+ .p2align 3; \
+ .word 4; \
+ .word 16; \
+ .word 5; \
+ .asciz "GNU"; \
+ .word type; \
+ .word 4; \
+ .word value; \
+ .word 0; \
+ .text
+
+#ifndef WANT_GNU_PROPERTY
+#define WANT_GNU_PROPERTY 1
+#endif
+
+#if WANT_GNU_PROPERTY
+/* Add property note with supported features to all asm files. */
+GNU_PROPERTY(FEATURE_1_AND, FEATURE_1_BTI|FEATURE_1_PAC)
+#endif
+
+#define ENTRY_ALIGN(name, alignment) \
+ .global name; \
+ .type name, %function; \
+ .align alignment; \
+name: \
+ .cfi_startproc; \
+ BTI_C;
+
+#define ENTRY(name) ENTRY_ALIGN(name, 6)
+
+#define ENTRY_ALIAS(name) \
+ .global name; \
+ .type name, %function; \
+ name:
+
+#define END(name) \
+ .cfi_endproc; \
+ .size name, .-name;
+
+#define L(l) .L ## l
+
+#endif
diff --git a/net/oenetcls/memcpy-sve.S b/net/oenetcls/memcpy-sve.S
new file mode 100644
index 000000000000..106e4c30294c
--- /dev/null
+++ b/net/oenetcls/memcpy-sve.S
@@ -0,0 +1,157 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#include "asmdefs.h"
+
+.arch armv8-a+sve
+
+#define dstin x0
+#define src x1
+#define count x2
+#define dst x3
+#define srcend x4
+#define dstend x5
+#define tmp1 x6
+#define vlen x6
+
+#define A_q q0
+#define B_q q1
+#define C_q q2
+#define D_q q3
+#define E_q q4
+#define F_q q5
+#define G_q q6
+#define H_q q7
+
+/* This implementation handles overlaps and supports both memcpy and memmove
+ from a single entry point. It uses unaligned accesses and branchless
+ sequences to keep the code small, simple and improve performance.
+ SVE vectors are used to speedup small copies.
+
+ Copies are split into 3 main cases: small copies of up to 32 bytes, medium
+ copies of up to 128 bytes, and large copies. The overhead of the overlap
+ check is negligible since it is only required for large copies.
+
+ Large copies use a software pipelined loop processing 64 bytes per iteration.
+ The source pointer is 16-byte aligned to minimize unaligned accesses.
+ The loop tail is handled by always copying 64 bytes from the end.
+*/
+
+ENTRY_ALIAS (__memmove_aarch64_sve)
+ENTRY (__memcpy_aarch64_sve)
+ cmp count, 128
+ b.hi L(copy_long)
+ cntb vlen
+ cmp count, vlen, lsl 1
+ b.hi L(copy32_128)
+
+ whilelo p0.b, xzr, count
+ whilelo p1.b, vlen, count
+ ld1b z0.b, p0/z, [src, 0, mul vl]
+ ld1b z1.b, p1/z, [src, 1, mul vl]
+ st1b z0.b, p0, [dstin, 0, mul vl]
+ st1b z1.b, p1, [dstin, 1, mul vl]
+ ret
+
+ /* Medium copies: 33..128 bytes. */
+L(copy32_128):
+ add srcend, src, count
+ add dstend, dstin, count
+ ldp A_q, B_q, [src]
+ ldp C_q, D_q, [srcend, -32]
+ cmp count, 64
+ b.hi L(copy128)
+ stp A_q, B_q, [dstin]
+ stp C_q, D_q, [dstend, -32]
+ ret
+
+ /* Copy 65..128 bytes. */
+L(copy128):
+ ldp E_q, F_q, [src, 32]
+ cmp count, 96
+ b.ls L(copy96)
+ ldp G_q, H_q, [srcend, -64]
+ stp G_q, H_q, [dstend, -64]
+L(copy96):
+ stp A_q, B_q, [dstin]
+ stp E_q, F_q, [dstin, 32]
+ stp C_q, D_q, [dstend, -32]
+ ret
+
+ /* Copy more than 128 bytes. */
+L(copy_long):
+ add srcend, src, count
+ add dstend, dstin, count
+
+ /* Use backwards copy if there is an overlap. */
+ sub tmp1, dstin, src
+ cmp tmp1, count
+ b.lo L(copy_long_backwards)
+
+ /* Copy 16 bytes and then align src to 16-byte alignment. */
+ ldr D_q, [src]
+ and tmp1, src, 15
+ bic src, src, 15
+ sub dst, dstin, tmp1
+ add count, count, tmp1 /* Count is now 16 too large. */
+ ldp A_q, B_q, [src, 16]
+ str D_q, [dstin]
+ ldp C_q, D_q, [src, 48]
+ subs count, count, 128 + 16 /* Test and readjust count. */
+ b.ls L(copy64_from_end)
+L(loop64):
+ stp A_q, B_q, [dst, 16]
+ ldp A_q, B_q, [src, 80]
+ stp C_q, D_q, [dst, 48]
+ ldp C_q, D_q, [src, 112]
+ add src, src, 64
+ add dst, dst, 64
+ subs count, count, 64
+ b.hi L(loop64)
+
+ /* Write the last iteration and copy 64 bytes from the end. */
+L(copy64_from_end):
+ ldp E_q, F_q, [srcend, -64]
+ stp A_q, B_q, [dst, 16]
+ ldp A_q, B_q, [srcend, -32]
+ stp C_q, D_q, [dst, 48]
+ stp E_q, F_q, [dstend, -64]
+ stp A_q, B_q, [dstend, -32]
+ ret
+
+ /* Large backwards copy for overlapping copies.
+ Copy 16 bytes and then align srcend to 16-byte alignment. */
+L(copy_long_backwards):
+ cbz tmp1, L(return)
+ ldr D_q, [srcend, -16]
+ and tmp1, srcend, 15
+ bic srcend, srcend, 15
+ sub count, count, tmp1
+ ldp A_q, B_q, [srcend, -32]
+ str D_q, [dstend, -16]
+ ldp C_q, D_q, [srcend, -64]
+ sub dstend, dstend, tmp1
+ subs count, count, 128
+ b.ls L(copy64_from_start)
+
+L(loop64_backwards):
+ str B_q, [dstend, -16]
+ str A_q, [dstend, -32]
+ ldp A_q, B_q, [srcend, -96]
+ str D_q, [dstend, -48]
+ str C_q, [dstend, -64]!
+ ldp C_q, D_q, [srcend, -128]
+ sub srcend, srcend, 64
+ subs count, count, 64
+ b.hi L(loop64_backwards)
+
+ /* Write the last iteration and copy 64 bytes from the start. */
+L(copy64_from_start):
+ ldp E_q, F_q, [src, 32]
+ stp A_q, B_q, [dstend, -32]
+ ldp A_q, B_q, [src]
+ stp C_q, D_q, [dstend, -64]
+ stp E_q, F_q, [dstin, 32]
+ stp A_q, B_q, [dstin]
+L(return):
+ ret
+
+END (__memcpy_aarch64_sve)
diff --git a/net/oenetcls/oenetcls.h b/net/oenetcls/oenetcls.h
new file mode 100644
index 000000000000..72aeedcaf0a0
--- /dev/null
+++ b/net/oenetcls/oenetcls.h
@@ -0,0 +1,176 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef _NET_OENETCLS_H
+#define _NET_OENETCLS_H
+#include <linux/if.h>
+#include <linux/mutex.h>
+#include <linux/cpufeature.h>
+
+#define OECLS_MAX_NETDEV_NUM 8
+#define OECLS_MAX_RXQ_NUM_PER_DEV 256
+#define OECLS_MAX_NUMA_NUM 16
+#define OECLS_MAX_CPU_NUM 1024
+
+#define OECLS_TIMEOUT (5 * HZ)
+#define OECLS_NO_FILTER 0xffff
+#define OECLS_NO_CPU 0xffff
+
+struct oecls_netdev_queue_info {
+ int irq;
+ int affinity_cpu;
+};
+
+struct oecls_netdev_info {
+ char dev_name[IFNAMSIZ];
+ struct net_device *netdev;
+ int rxq_num;
+ struct oecls_netdev_queue_info rxq[OECLS_MAX_RXQ_NUM_PER_DEV];
+ int old_filter_state;
+};
+
+struct oecls_numa_bound_dev_info {
+ DECLARE_BITMAP(bitmap_rxq, OECLS_MAX_RXQ_NUM_PER_DEV);
+};
+
+struct oecls_numa_info {
+ DECLARE_BITMAP(avail_cpus, OECLS_MAX_CPU_NUM);
+ struct oecls_numa_bound_dev_info bound_dev[OECLS_MAX_NETDEV_NUM];
+};
+
+struct cmd_context {
+ char netdev[IFNAMSIZ];
+ u32 dip4;
+ u16 dport;
+ u16 action;
+ u32 ruleid;
+ u32 del_ruleid;
+ int ret_loc;
+};
+
+#define OECLS_SK_RULE_HASHSIZE 256
+#define OECLS_SK_RULE_HASHMASK (OECLS_SK_RULE_HASHSIZE - 1)
+
+struct oecls_sk_rule_list {
+ struct hlist_head hash[OECLS_SK_RULE_HASHSIZE];
+ /* Mutex to synchronize access to ntuple rule locking */
+ struct mutex mutex;
+};
+
+struct oecls_sk_rule {
+ struct hlist_node node;
+ int devid;
+ void *sk;
+ int dip4;
+ int dport;
+ int action;
+ int ruleid;
+ int nid;
+};
+
+struct oecls_dev_flow {
+ unsigned short cpu;
+ unsigned short filter;
+ unsigned int last_qtail;
+ int isvalid;
+ unsigned long timeout;
+};
+
+struct oecls_dev_flow_table {
+ unsigned int mask;
+ struct rcu_head rcu;
+ struct oecls_dev_flow flows[];
+};
+
+struct oecls_sock_flow_table {
+ u32 mask;
+ u32 ents[] ____cacheline_aligned_in_smp;
+};
+
+#define OECLS_DEV_FLOW_TABLE_NUM 0x1000
+#define OECLS_SOCK_FLOW_TABLE_NUM 0x100000
+#define OECLS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct oecls_dev_flow_table) + \
+ ((_num) * sizeof(struct oecls_dev_flow)))
+#define OECLS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct oecls_sock_flow_table, ents[_num]))
+
+#define ETH_ALL_FLAGS (ETH_FLAG_LRO | ETH_FLAG_RXVLAN | ETH_FLAG_TXVLAN | \
+ ETH_FLAG_NTUPLE | ETH_FLAG_RXHASH)
+#define ETH_ALL_FEATURES (NETIF_F_LRO | NETIF_F_HW_VLAN_CTAG_RX | \
+ NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_NTUPLE | \
+ NETIF_F_RXHASH)
+
+struct rmgr_ctrl {
+ int driver_select;
+ unsigned long *slot;
+ __u32 n_rules;
+ __u32 size;
+};
+
+extern int match_ip_flag;
+extern int debug;
+extern int oecls_netdev_num;
+extern int oecls_numa_num;
+
+#define oecls_debug(fmt, ...) \
+ do { \
+ if (debug) \
+ trace_printk(fmt, ## __VA_ARGS__); \
+ } while (0)
+
+#define oecls_error(fmt, ...) \
+ do { \
+ pr_err("oenetcls [%s:%d]: " fmt, __FILE__, __LINE__, ## __VA_ARGS__); \
+ trace_printk(fmt, ## __VA_ARGS__); \
+ } while (0)
+
+struct oecls_netdev_info *get_oecls_netdev_info(unsigned int index);
+
+#define for_each_oecls_netdev(devid, oecls_dev) \
+ for (devid = 0, oecls_dev = get_oecls_netdev_info(devid); \
+ (devid < oecls_netdev_num) && oecls_dev; \
+ devid++, oecls_dev = get_oecls_netdev_info(devid))
+
+struct oecls_numa_info *get_oecls_numa_info(unsigned int nid);
+
+#define for_each_oecls_numa(nid, numa_info) \
+ for (nid = 0, numa_info = get_oecls_numa_info(nid); \
+ (nid < oecls_numa_num) && numa_info; \
+ nid++, numa_info = get_oecls_numa_info(nid))
+
+#ifdef CONFIG_ARM64_SVE
+void *__memcpy_aarch64_sve(void *, const void *, size_t);
+#define memcpy_r(dst, src, len) \
+ do { \
+ if (system_supports_sve()) \
+ __memcpy_aarch64_sve(dst, src, len); \
+ else \
+ memcpy(dst, src, len); \
+ } while (0)
+#else
+#define memcpy_r(dst, src, len) memcpy(dst, src, len)
+#endif
+
+int check_appname(char *task_name);
+int send_ethtool_ioctl(struct cmd_context *ctx, void *cmd);
+int alloc_rxq_id(int nid, int devid);
+void free_rxq_id(int nid, int devid, int rxq_id);
+void oecls_ntuple_res_init(void);
+void oecls_ntuple_res_clean(void);
+void oecls_flow_res_init(void);
+void oecls_flow_res_clean(void);
+
+#define L0_MAX_PAGE_SIZE (8192)
+#define L0_MAX_PAGE_NUM (4096)
+
+struct l0_vma_data {
+ struct page *page;
+ unsigned long size;
+ int nid;
+};
+
+void clean_oecls_l0_cache(void);
+void init_oecls_l0_cache(void);
+void *alloc_from_l0(int size);
+void free_to_l0(void *addr);
+int l3t_shared_lock(int nid, unsigned long pfn, unsigned long size);
+int l3t_shared_unlock(int nid, unsigned long pfn, unsigned long size);
+
+#endif /* _NET_OENETCLS_H */
diff --git a/net/oenetcls/oenetcls_flow.c b/net/oenetcls/oenetcls_flow.c
new file mode 100644
index 000000000000..c35ce5aca416
--- /dev/null
+++ b/net/oenetcls/oenetcls_flow.c
@@ -0,0 +1,402 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/inetdevice.h>
+#include <linux/netdevice.h>
+#include <linux/rtnetlink.h>
+#include <linux/irq.h>
+#include <linux/irqdesc.h>
+#include <linux/inet.h>
+#include <net/sock.h>
+#include <trace/hooks/oenetcls.h>
+#include "oenetcls.h"
+
+static u32 oecls_cpu_mask;
+static struct oecls_sock_flow_table __rcu *oecls_sock_flow_table;
+static DEFINE_MUTEX(oecls_sock_flow_mutex);
+static DEFINE_SPINLOCK(oecls_dev_flow_lock);
+
+bool is_oecls_config_netdev(const char *name)
+{
+ struct oecls_netdev_info *netdev_info;
+ int netdev_loop;
+
+ for_each_oecls_netdev(netdev_loop, netdev_info)
+ if (strcmp(netdev_info->dev_name, name) == 0)
+ return true;
+
+ return false;
+}
+
+static void oecls_timeout(void *data, struct net_device *dev, u16 rxq_index,
+ u32 flow_id, u16 filter_id, bool *ret)
+{
+ struct netdev_rx_queue *rxqueue = dev->_rx + rxq_index;
+ struct oecls_dev_flow_table *flow_table;
+ struct oecls_dev_flow *rflow;
+ bool expire = true;
+ unsigned int cpu;
+
+ rcu_read_lock();
+ flow_table = rcu_dereference(rxqueue->oecls_ftb);
+ if (flow_table && flow_id <= flow_table->mask) {
+ rflow = &flow_table->flows[flow_id];
+ cpu = READ_ONCE(rflow->cpu);
+ oecls_debug("dev:%s, rxq:%d, flow_id:%u, filter_id:%d/%d, cpu:%d", dev->name,
+ rxq_index, flow_id, filter_id, rflow->filter, cpu);
+
+ if (rflow->filter == filter_id && cpu < nr_cpu_ids) {
+ if (time_before(jiffies, rflow->timeout + OECLS_TIMEOUT)) {
+ expire = false;
+ } else {
+ rflow->isvalid = 0;
+ WRITE_ONCE(rflow->cpu, OECLS_NO_CPU);
+ }
+ }
+ }
+ rcu_read_unlock();
+ oecls_debug("%s, dev:%s, rxq:%d, flow_id:%u, filter_id:%d, expire:%d\n", __func__,
+ dev->name, rxq_index, flow_id, filter_id, expire);
+ *ret = expire;
+}
+
+static void oecls_flow_update(void *data, struct sock *sk)
+{
+ struct oecls_sock_flow_table *tb;
+ unsigned int hash, index;
+ u32 val;
+ u32 cpu = raw_smp_processor_id();
+
+ if (sk->sk_state != TCP_ESTABLISHED)
+ return;
+
+ if (check_appname(current->comm))
+ return;
+
+ rcu_read_lock();
+ tb = rcu_dereference(oecls_sock_flow_table);
+ hash = READ_ONCE(sk->sk_rxhash);
+ if (tb && hash) {
+ index = hash & tb->mask;
+ val = hash & ~oecls_cpu_mask;
+ val |= cpu;
+
+ if (READ_ONCE(tb->ents[index]) != val) {
+ WRITE_ONCE(tb->ents[index], val);
+
+ oecls_debug("[%s] sk:%p, hash:0x%x, index:0x%x, val:0x%x, cpu:%d\n",
+ current->comm, sk, hash, index, val, cpu);
+ }
+ }
+ rcu_read_unlock();
+}
+
+static int flow_get_queue_idx(struct net_device *dev, int nid, struct sk_buff *skb)
+{
+ struct oecls_netdev_info *netdev_info;
+ int netdev_loop;
+ u32 hash, index;
+ struct oecls_numa_info *numa_info;
+ struct oecls_numa_bound_dev_info *bound_dev = NULL;
+ int rxq_id, rxq_num, i;
+
+ numa_info = get_oecls_numa_info(nid);
+ if (!numa_info)
+ return -1;
+
+ for_each_oecls_netdev(netdev_loop, netdev_info) {
+ if (strcmp(netdev_info->dev_name, dev->name) == 0) {
+ bound_dev = &numa_info->bound_dev[netdev_loop];
+ break;
+ }
+ }
+
+ if (!bound_dev)
+ return -1;
+ rxq_num = bitmap_weight(bound_dev->bitmap_rxq, OECLS_MAX_RXQ_NUM_PER_DEV);
+ if (rxq_num == 0)
+ return -1;
+
+ hash = skb_get_hash(skb);
+ index = hash % rxq_num;
+
+ i = 0;
+ for_each_set_bit(rxq_id, bound_dev->bitmap_rxq, OECLS_MAX_RXQ_NUM_PER_DEV)
+ if (index == i++)
+ return rxq_id;
+
+ return -1;
+}
+
+static void set_oecls_cpu(struct net_device *dev, struct sk_buff *skb,
+ struct oecls_dev_flow *old_rflow, int old_rxq_id, u16 next_cpu)
+{
+ struct netdev_rx_queue *rxqueue;
+ struct oecls_dev_flow_table *dtb;
+ struct oecls_dev_flow *rflow;
+ u32 flow_id, hash;
+ u16 rxq_index;
+ int rc;
+
+ if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
+ !(dev->features & NETIF_F_NTUPLE))
+ return;
+
+ rxq_index = flow_get_queue_idx(dev, cpu_to_node(next_cpu), skb);
+ if (rxq_index == skb_get_rx_queue(skb) || rxq_index < 0)
+ return;
+
+ rxqueue = dev->_rx + rxq_index;
+ dtb = rcu_dereference(rxqueue->oecls_ftb);
+ if (!dtb)
+ return;
+
+ hash = skb_get_hash(skb);
+ flow_id = hash & dtb->mask;
+ rflow = &dtb->flows[flow_id];
+ if (rflow->isvalid && rflow->cpu == next_cpu) {
+ rflow->timeout = jiffies;
+ return;
+ }
+
+ rc = dev->netdev_ops->ndo_rx_flow_steer(dev, skb, rxq_index, flow_id);
+ oecls_debug("skb:%p, rxq:%d, hash:0x%x, flow_id:%u, old_rxq_id:%d, next_cpu:%d, rc:%d\n",
+ skb, rxq_index, hash, flow_id, old_rxq_id, next_cpu, rc);
+ if (rc < 0)
+ return;
+
+ rflow->filter = rc;
+ rflow->isvalid = 1;
+ rflow->timeout = jiffies;
+ if (old_rflow->filter == rflow->filter)
+ old_rflow->filter = OECLS_NO_FILTER;
+ rflow->cpu = next_cpu;
+}
+
+static void __oecls_set_cpu(struct sk_buff *skb, struct net_device *ndev,
+ struct oecls_sock_flow_table *tb, struct oecls_dev_flow_table *dtb,
+ int old_rxq_id)
+{
+ struct oecls_dev_flow *rflow;
+ u32 last_recv_cpu, hash, val;
+ u32 tcpu = 0;
+ u32 cpu = raw_smp_processor_id();
+
+ skb_reset_network_header(skb);
+ hash = skb_get_hash(skb);
+ if (!hash)
+ return;
+
+ val = READ_ONCE(tb->ents[hash & tb->mask]);
+ last_recv_cpu = val & oecls_cpu_mask;
+ rflow = &dtb->flows[hash & dtb->mask];
+ tcpu = rflow->cpu;
+
+ if ((val ^ hash) & ~oecls_cpu_mask)
+ return;
+
+ if (cpu_to_node(cpu) == cpu_to_node(last_recv_cpu))
+ return;
+
+ if (tcpu >= nr_cpu_ids)
+ set_oecls_cpu(ndev, skb, rflow, old_rxq_id, last_recv_cpu);
+}
+
+static void oecls_set_cpu(void *data, struct sk_buff *skb)
+{
+ struct net_device *ndev = skb->dev;
+ struct oecls_sock_flow_table *stb;
+ struct oecls_dev_flow_table *dtb;
+ struct netdev_rx_queue *rxqueue;
+ int rxq_id = -1;
+
+ if (!ndev)
+ return;
+
+ if (!is_oecls_config_netdev(ndev->name))
+ return;
+
+ rxqueue = ndev->_rx;
+ if (skb_rx_queue_recorded(skb)) {
+ rxq_id = skb_get_rx_queue(skb);
+ if (rxq_id >= ndev->real_num_rx_queues) {
+ oecls_debug("ndev:%s, rxq:%d, real_num:%d\n", ndev->name,
+ rxq_id, ndev->real_num_rx_queues);
+ return;
+ }
+ rxqueue += rxq_id;
+ }
+
+ // oecls_debug("skb:%px, dev:%s, rxq_id:%d\n", skb, ndev->name, rxq_id);
+ if (rxq_id < 0)
+ return;
+
+ rcu_read_lock();
+ stb = rcu_dereference(oecls_sock_flow_table);
+ dtb = rcu_dereference(rxqueue->oecls_ftb);
+ if (stb && dtb)
+ __oecls_set_cpu(skb, ndev, stb, dtb, rxq_id);
+
+ rcu_read_unlock();
+}
+
+static void oecls_dev_flow_table_free(struct rcu_head *rcu)
+{
+ struct oecls_dev_flow_table *table = container_of(rcu,
+ struct oecls_dev_flow_table, rcu);
+ vfree(table);
+}
+
+static void oecls_dev_flow_table_cleanup(struct net_device *netdev, int qid)
+{
+ struct oecls_dev_flow_table *dtb;
+ struct netdev_rx_queue *queue;
+ int i;
+
+ spin_lock(&oecls_dev_flow_lock);
+ for (i = 0; i < qid; i++) {
+ queue = netdev->_rx + i;
+ dtb = rcu_dereference_protected(queue->oecls_ftb,
+ lockdep_is_held(&oecls_dev_flow_lock));
+ rcu_assign_pointer(queue->oecls_ftb, NULL);
+ }
+ spin_unlock(&oecls_dev_flow_lock);
+ call_rcu(&dtb->rcu, oecls_dev_flow_table_free);
+}
+
+static int oecls_dev_flow_table_release(void)
+{
+ struct oecls_netdev_info *netdev_info;
+ int netdev_loop;
+ struct net_device *netdev;
+
+ for_each_oecls_netdev(netdev_loop, netdev_info) {
+ netdev = netdev_info->netdev;
+ if (!netdev)
+ continue;
+ oecls_dev_flow_table_cleanup(netdev, netdev->num_rx_queues);
+ }
+
+ return 0;
+}
+
+static int _oecls_dev_flow_table_init(struct net_device *netdev)
+{
+ struct oecls_dev_flow_table *table;
+ int size = OECLS_DEV_FLOW_TABLE_NUM;
+ struct netdev_rx_queue *queue;
+ int i, j, ret = 0;
+
+ size = roundup_pow_of_two(size);
+ oecls_debug("dev:%s, num_rx_queues:%d, mask:0x%x\n", netdev->name, netdev->num_rx_queues,
+ size - 1);
+
+ for (i = 0; i < netdev->num_rx_queues; i++) {
+ table = vmalloc(OECLS_DEV_FLOW_TABLE_SIZE(size));
+ if (!table) {
+ ret = -ENOMEM;
+ goto fail;
+ }
+
+ table->mask = size - 1;
+ for (j = 0; j < size; j++) {
+ table->flows[j].cpu = OECLS_NO_CPU;
+ table->flows[j].isvalid = 0;
+ }
+
+ queue = netdev->_rx + i;
+
+ spin_lock(&oecls_dev_flow_lock);
+ rcu_assign_pointer(queue->oecls_ftb, table);
+ spin_unlock(&oecls_dev_flow_lock);
+ }
+ return ret;
+fail:
+ oecls_dev_flow_table_cleanup(netdev, i);
+ return ret;
+}
+
+static int oecls_dev_flow_table_init(void)
+{
+ struct oecls_netdev_info *netdev_info;
+ int netdev_loop;
+ struct net_device *ndev;
+ int i, err;
+
+ for_each_oecls_netdev(netdev_loop, netdev_info) {
+ ndev = netdev_info->netdev;
+ if (!ndev)
+ continue;
+ err = _oecls_dev_flow_table_init(ndev);
+ if (err)
+ goto out;
+ }
+
+ return 0;
+out:
+ for (i = 0; i < netdev_loop; i++) {
+ netdev_info = get_oecls_netdev_info(i);
+ ndev = netdev_info->netdev;
+ if (!ndev)
+ continue;
+ oecls_dev_flow_table_cleanup(ndev, ndev->num_rx_queues);
+ }
+ return err;
+}
+
+static int oecls_sock_flow_table_release(void)
+{
+ struct oecls_sock_flow_table *tb;
+
+ mutex_lock(&oecls_sock_flow_mutex);
+ tb = rcu_dereference_protected(oecls_sock_flow_table,
+ lockdep_is_held(&oecls_sock_flow_mutex));
+ if (tb)
+ rcu_assign_pointer(oecls_sock_flow_table, NULL);
+ mutex_unlock(&oecls_sock_flow_mutex);
+ synchronize_rcu();
+ vfree(tb);
+
+ unregister_trace_oecls_flow_update(&oecls_flow_update, NULL);
+ unregister_trace_oecls_set_cpu(&oecls_set_cpu, NULL);
+ unregister_trace_oecls_timeout(&oecls_timeout, NULL);
+ return 0;
+}
+
+static int oecls_sock_flow_table_init(void)
+{
+ struct oecls_sock_flow_table *table;
+ int size = OECLS_SOCK_FLOW_TABLE_NUM;
+ int i;
+
+ size = roundup_pow_of_two(size);
+ table = vmalloc(OECLS_SOCK_FLOW_TABLE_SIZE(size));
+ if (!table)
+ return -ENOMEM;
+
+ oecls_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1;
+ oecls_debug("nr_cpu_ids:%d, oecls_cpu_mask:0x%x\n", nr_cpu_ids, oecls_cpu_mask);
+
+ table->mask = size - 1;
+ for (i = 0; i < size; i++)
+ table->ents[i] = OECLS_NO_CPU;
+
+ mutex_lock(&oecls_sock_flow_mutex);
+ rcu_assign_pointer(oecls_sock_flow_table, table);
+ mutex_unlock(&oecls_sock_flow_mutex);
+
+ register_trace_oecls_flow_update(oecls_flow_update, NULL);
+ register_trace_oecls_set_cpu(&oecls_set_cpu, NULL);
+ register_trace_oecls_timeout(&oecls_timeout, NULL);
+ return 0;
+}
+
+void oecls_flow_res_init(void)
+{
+ oecls_sock_flow_table_init();
+ oecls_dev_flow_table_init();
+}
+
+void oecls_flow_res_clean(void)
+{
+ oecls_sock_flow_table_release();
+ oecls_dev_flow_table_release();
+}
diff --git a/net/oenetcls/oenetcls_l0.c b/net/oenetcls/oenetcls_l0.c
new file mode 100644
index 000000000000..b820a1cf96bd
--- /dev/null
+++ b/net/oenetcls/oenetcls_l0.c
@@ -0,0 +1,193 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/fs.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/miscdevice.h>
+#include <linux/kallsyms.h>
+#include <linux/mm.h>
+#include <linux/gfp.h>
+#include <linux/mempolicy.h>
+#include <linux/pfn_t.h>
+#include <linux/delay.h>
+#include "oenetcls.h"
+
+static struct vm_area_struct *oecls_l0_cache __ro_after_init;
+static void *oecls_l0_cache_head;
+static atomic_t alloc_num_l0 = ATOMIC_INIT(0);
+static int cur_offset_l0;
+static u8 slot_l0[L0_MAX_PAGE_NUM] = { 0 };
+
+void *alloc_from_l0(int size)
+{
+ int i;
+ void *addr = NULL;
+
+ if (!oecls_l0_cache_head)
+ return kzalloc(size, GFP_ATOMIC);
+
+ if (size > L0_MAX_PAGE_SIZE) {
+ oecls_error("size(%d) >= max_size(%d)!\n", size, L0_MAX_PAGE_SIZE);
+ return NULL;
+ }
+
+ if (atomic_read(&alloc_num_l0) >= L0_MAX_PAGE_NUM) {
+ oecls_error("alloc_num_l0:%d exceed max num!\n", atomic_read(&alloc_num_l0));
+ return NULL;
+ }
+
+ for (i = 0; i < L0_MAX_PAGE_NUM; i++) {
+ cur_offset_l0 = (cur_offset_l0 + 1) % L0_MAX_PAGE_NUM;
+ if (slot_l0[cur_offset_l0] == 0) {
+ addr = oecls_l0_cache_head + cur_offset_l0 * L0_MAX_PAGE_SIZE;
+ atomic_inc(&alloc_num_l0);
+ slot_l0[cur_offset_l0] = 1;
+ break;
+ }
+ }
+
+ if (!addr)
+ oecls_error("app:%s, pid:%d alloc fail!\n", current->comm, current->pid);
+ return addr;
+}
+
+void free_to_l0(void *addr)
+{
+ int offset = 0;
+
+ if (!addr)
+ return;
+
+ if (!oecls_l0_cache_head) {
+ kfree(addr);
+ return;
+ }
+
+ if (atomic_read(&alloc_num_l0) <= 0) {
+ oecls_error("alloc_num_l0:%d <= 0!\n", atomic_read(&alloc_num_l0));
+ return;
+ }
+
+ offset = (addr - oecls_l0_cache_head) / L0_MAX_PAGE_SIZE;
+ if (offset >= L0_MAX_PAGE_NUM) {
+ oecls_error("app:%s, pid:%d, offset:%d\n", current->comm, current->pid, offset);
+ return;
+ }
+
+ slot_l0[offset] = 0;
+ atomic_dec(&alloc_num_l0);
+}
+
+static int get_node_node(struct vm_area_struct *vma)
+{
+ struct mempolicy *pol;
+ nodemask_t *nmask;
+ int nid;
+
+ nid = get_vma_policy_node(vma, vma->vm_start, GFP_KERNEL, &pol, &nmask);
+ if (pol->mode == MPOL_BIND || pol->mode == MPOL_PREFERRED_MANY)
+ nid = first_node(*nmask);
+
+ return nid;
+}
+
+static int oecls_l0_cache_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+ unsigned long cont_size = PAGE_ALIGN(vma->vm_end - vma->vm_start);
+ struct page *page = NULL;
+ struct l0_vma_data *data;
+ int page_cnt, nid;
+ int ret = 0;
+
+ if ((vma->vm_start % PMD_SIZE) || (vma->vm_end % PMD_SIZE)) {
+ oecls_error("error vma_start: %#lx, vma_end: %#lx\n", vma->vm_start, vma->vm_end);
+ return -EINVAL;
+ }
+
+ data = kzalloc(sizeof(*data), GFP_KERNEL);
+ if (!data) {
+ oecls_error("kzalloc l0_vma_data fail!\n");
+ return -ENOMEM;
+ }
+
+ page_cnt = cont_size >> PAGE_SHIFT;
+ nid = get_node_node(vma);
+#ifdef CONFIG_CONTIG_ALLOC
+ page = alloc_contig_pages(page_cnt, GFP_KERNEL, nid, NULL);
+#endif
+ if (!page) {
+ ret = -ENOMEM;
+ oecls_error("alloc_contig_pages fail! page_cnt:%d, nid:%d\n", page_cnt, nid);
+ goto free_data;
+ }
+
+#if IS_ENABLED(CONFIG_HISI_L3T)
+ ret = l3t_shared_lock(nid, page_to_pfn(page), cont_size);
+#endif
+ if (ret) {
+ ret = -EOPNOTSUPP;
+ oecls_error("l3t_shared_lock fail! ret: %d\n", ret);
+ goto free_page;
+ }
+
+ data->page = page;
+ data->size = cont_size;
+ data->nid = nid;
+ vma->vm_private_data = data;
+ return 0;
+free_page:
+ free_contig_range(page_to_pfn(page), page_cnt);
+free_data:
+ kfree(data);
+ return ret;
+}
+
+void init_oecls_l0_cache(void)
+{
+ int ret = 0;
+ struct l0_vma_data *sbkpages;
+
+ oecls_l0_cache = kzalloc(sizeof(*oecls_l0_cache), GFP_KERNEL);
+ if (!oecls_l0_cache) {
+ oecls_error("kzalloc oecls_l0_cache fail!\n");
+ return;
+ }
+ oecls_l0_cache->vm_start = 0;
+ oecls_l0_cache->vm_end = L0_MAX_PAGE_NUM * L0_MAX_PAGE_SIZE;
+
+ ret = oecls_l0_cache_mmap(NULL, oecls_l0_cache);
+ if (ret != 0) {
+ kfree(oecls_l0_cache);
+ oecls_l0_cache = NULL;
+ return;
+ }
+ sbkpages = (struct l0_vma_data *)(oecls_l0_cache->vm_private_data);
+ oecls_l0_cache_head = page_to_virt(sbkpages->page);
+
+ atomic_set(&alloc_num_l0, 0);
+ oecls_debug("l0_cache:%p, l0_cache_head:%p\n", oecls_l0_cache, oecls_l0_cache_head);
+}
+
+void clean_oecls_l0_cache(void)
+{
+ struct l0_vma_data *sbkpages;
+
+ if (!oecls_l0_cache)
+ return;
+
+ oecls_debug("alloc_num_l0:%d\n", atomic_read(&alloc_num_l0));
+ while (atomic_read(&alloc_num_l0) != 0)
+ mdelay(1);
+
+ sbkpages = (struct l0_vma_data *)(oecls_l0_cache->vm_private_data);
+ if (sbkpages) {
+#if IS_ENABLED(CONFIG_HISI_L3T)
+ l3t_shared_unlock(sbkpages->nid, page_to_pfn(sbkpages->page), sbkpages->size);
+#endif
+ free_contig_range(page_to_pfn(sbkpages->page), sbkpages->size >> PAGE_SHIFT);
+ kfree(sbkpages);
+ oecls_l0_cache_head = NULL;
+ }
+
+ kfree(oecls_l0_cache);
+}
diff --git a/net/oenetcls/oenetcls_main.c b/net/oenetcls/oenetcls_main.c
new file mode 100644
index 000000000000..8cf5d02bb5b9
--- /dev/null
+++ b/net/oenetcls/oenetcls_main.c
@@ -0,0 +1,972 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/netdev_features.h>
+#include <linux/ethtool.h>
+#include <linux/irq.h>
+#include <linux/irqdesc.h>
+#include <linux/rtnetlink.h>
+#include "oenetcls.h"
+
+int oecls_netdev_num;
+static struct oecls_netdev_info oecls_netdev_info_table[OECLS_MAX_NETDEV_NUM];
+
+int oecls_numa_num;
+static struct oecls_numa_info oecls_numa_info_table[OECLS_MAX_NUMA_NUM];
+
+int debug;
+module_param(debug, int, 0644);
+MODULE_PARM_DESC(debug, "debug switch");
+
+static int mode;
+module_param(mode, int, 0444);
+MODULE_PARM_DESC(mode, "mode, default 0");
+
+static char ifname[64] = { 0 };
+module_param_string(ifname, ifname, sizeof(ifname), 0444);
+MODULE_PARM_DESC(ifname, "ifname");
+
+static char appname[64] = "redis-server";
+module_param_string(appname, appname, sizeof(appname), 0644);
+MODULE_PARM_DESC(appname, "appname, default redis-server");
+
+int match_ip_flag = 1;
+module_param(match_ip_flag, int, 0644);
+MODULE_PARM_DESC(match_ip_flag, "match ip flag");
+
+static int strategy;
+module_param(strategy, int, 0444);
+MODULE_PARM_DESC(strategy, "strategy, default 0");
+
+static bool check_params(void)
+{
+ if (mode != 0 && mode != 1)
+ return false;
+
+ if (strlen(ifname) == 0)
+ return false;
+
+ return true;
+}
+
+int check_appname(char *task_name)
+{
+ char *start = appname;
+ char *end;
+
+ if (!strlen(appname))
+ return 0;
+
+ // support appname: app1#app2#appN
+ while (*start != '\0') {
+ end = strchr(start, '#');
+ if (end == start) {
+ start++;
+ continue;
+ }
+
+ if (!end) {
+ if (!strncmp(task_name, start, strlen(start)))
+ return 0;
+ break;
+ }
+
+ if (!strncmp(task_name, start, end - start))
+ return 0;
+ start = end + 1;
+ }
+ return -EOPNOTSUPP;
+}
+
+static u32 __ethtool_get_flags(struct net_device *dev)
+{
+ u32 flags = 0;
+
+ if (dev->features & NETIF_F_LRO)
+ flags |= ETH_FLAG_LRO;
+ if (dev->features & NETIF_F_HW_VLAN_CTAG_RX)
+ flags |= ETH_FLAG_RXVLAN;
+ if (dev->features & NETIF_F_HW_VLAN_CTAG_TX)
+ flags |= ETH_FLAG_TXVLAN;
+ if (dev->features & NETIF_F_NTUPLE)
+ flags |= ETH_FLAG_NTUPLE;
+ if (dev->features & NETIF_F_RXHASH)
+ flags |= ETH_FLAG_RXHASH;
+
+ return flags;
+}
+
+static int __ethtool_set_flags(struct net_device *dev, u32 data)
+{
+ netdev_features_t features = 0, changed;
+
+ if (data & ~ETH_ALL_FLAGS)
+ return -EINVAL;
+
+ if (data & ETH_FLAG_LRO)
+ features |= NETIF_F_LRO;
+ if (data & ETH_FLAG_RXVLAN)
+ features |= NETIF_F_HW_VLAN_CTAG_RX;
+ if (data & ETH_FLAG_TXVLAN)
+ features |= NETIF_F_HW_VLAN_CTAG_TX;
+ if (data & ETH_FLAG_NTUPLE)
+ features |= NETIF_F_NTUPLE;
+ if (data & ETH_FLAG_RXHASH)
+ features |= NETIF_F_RXHASH;
+
+ /* allow changing only bits set in hw_features */
+ changed = (features ^ dev->features) & ETH_ALL_FEATURES;
+ if (changed & ~dev->hw_features)
+ return (changed & dev->hw_features) ? -EINVAL : -EOPNOTSUPP;
+
+ dev->wanted_features =
+ (dev->wanted_features & ~changed) | (features & changed);
+
+ __netdev_update_features(dev);
+
+ return 0;
+}
+
+static void ethtool_rxnfc_copy_to_user(void *useraddr,
+ const struct ethtool_rxnfc *rxnfc,
+ size_t size, const u32 *rule_buf)
+{
+ memcpy_r(useraddr, rxnfc, size);
+ useraddr += offsetof(struct ethtool_rxnfc, rule_locs);
+
+ if (rule_buf)
+ memcpy_r(useraddr, rule_buf, rxnfc->rule_cnt * sizeof(u32));
+}
+
+static noinline_for_stack int ethtool_set_rxnfc(struct net_device *dev,
+ u32 cmd, void *useraddr)
+{
+ struct ethtool_rxnfc info;
+ size_t info_size = sizeof(info);
+ int rc;
+
+ if (!dev->ethtool_ops->set_rxnfc)
+ return -EOPNOTSUPP;
+
+ if (cmd == ETHTOOL_SRXFH)
+ info_size = (offsetof(struct ethtool_rxnfc, data) +
+ sizeof(info.data));
+
+ memcpy_r(&info, useraddr, info_size);
+ rc = dev->ethtool_ops->set_rxnfc(dev, &info);
+ if (rc)
+ return rc;
+
+ if (cmd == ETHTOOL_SRXCLSRLINS)
+ ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, NULL);
+
+ return 0;
+}
+
+static noinline_for_stack int ethtool_get_rxnfc(struct net_device *dev,
+ u32 cmd, void *useraddr)
+{
+ struct ethtool_rxnfc info;
+ size_t info_size = sizeof(info);
+ const struct ethtool_ops *ops = dev->ethtool_ops;
+ int ret;
+ void *rule_buf = NULL;
+
+ if (!ops->get_rxnfc)
+ return -EOPNOTSUPP;
+
+ if (cmd == ETHTOOL_GRXFH)
+ info_size = (offsetof(struct ethtool_rxnfc, data) +
+ sizeof(info.data));
+
+ memcpy_r(&info, useraddr, info_size);
+
+ /* If FLOW_RSS was requested then user-space must be using the
+ * new definition, as FLOW_RSS is newer.
+ */
+ if (cmd == ETHTOOL_GRXFH && info.flow_type & FLOW_RSS) {
+ info_size = sizeof(info);
+ memcpy_r(&info, useraddr, info_size);
+ /* Since malicious users may modify the original data,
+ * we need to check whether FLOW_RSS is still requested.
+ */
+ if (!(info.flow_type & FLOW_RSS))
+ return -EINVAL;
+ }
+
+ if (info.cmd != cmd)
+ return -EINVAL;
+
+ if (info.cmd == ETHTOOL_GRXCLSRLALL) {
+ if (info.rule_cnt > 0) {
+ if (info.rule_cnt <= KMALLOC_MAX_SIZE / sizeof(u32))
+ rule_buf = kcalloc(info.rule_cnt, sizeof(u32),
+ GFP_KERNEL);
+ if (!rule_buf)
+ return -ENOMEM;
+ }
+ }
+
+ ret = ops->get_rxnfc(dev, &info, rule_buf);
+ if (ret < 0)
+ goto err_out;
+
+ ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, rule_buf);
+err_out:
+ kfree(rule_buf);
+
+ return ret;
+}
+
+static noinline_for_stack int ethtool_get_channels(struct net_device *dev,
+ void *useraddr)
+{
+ struct ethtool_channels channels = { .cmd = ETHTOOL_GCHANNELS };
+
+ if (!dev->ethtool_ops->get_channels)
+ return -EOPNOTSUPP;
+
+ dev->ethtool_ops->get_channels(dev, &channels);
+
+ memcpy_r(useraddr, &channels, sizeof(channels));
+ return 0;
+}
+
+static int ethtool_get_value(struct net_device *dev, char *useraddr,
+ u32 cmd, u32 (*actor)(struct net_device *))
+{
+ struct ethtool_value edata = { .cmd = cmd };
+
+ if (!actor)
+ return -EOPNOTSUPP;
+
+ edata.data = actor(dev);
+
+ memcpy_r(useraddr, &edata, sizeof(edata));
+ return 0;
+}
+
+static int ethtool_set_value(struct net_device *dev, char *useraddr,
+ int (*actor)(struct net_device *, u32))
+{
+ struct ethtool_value edata;
+
+ if (!actor)
+ return -EOPNOTSUPP;
+
+ memcpy_r(&edata, useraddr, sizeof(edata));
+
+ return actor(dev, edata.data);
+}
+
+static int dev_ethtool_kern(struct net *net, struct ifreq *ifr)
+{
+ struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name);
+ void *useraddr = ifr->ifr_data;
+ u32 ethcmd, sub_cmd;
+ int rc;
+ netdev_features_t old_features;
+
+ if (!dev || !netif_device_present(dev))
+ return -ENODEV;
+
+ memcpy_r(ðcmd, useraddr, sizeof(ethcmd));
+
+ if (ethcmd == ETHTOOL_PERQUEUE)
+ memcpy_r(&sub_cmd, useraddr + sizeof(ethcmd), sizeof(sub_cmd));
+ else
+ sub_cmd = ethcmd;
+
+ /* Allow some commands to be done by anyone */
+ switch (sub_cmd) {
+ case ETHTOOL_GFLAGS:
+ case ETHTOOL_GRXFH:
+ case ETHTOOL_GRXRINGS:
+ case ETHTOOL_GRXCLSRLCNT:
+ case ETHTOOL_GRXCLSRULE:
+ case ETHTOOL_GRXCLSRLALL:
+ case ETHTOOL_GCHANNELS:
+ break;
+ default:
+ if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
+ return -EPERM;
+ }
+
+ if (dev->ethtool_ops->begin) {
+ rc = dev->ethtool_ops->begin(dev);
+ if (rc < 0)
+ return rc;
+ }
+ old_features = dev->features;
+
+ switch (ethcmd) {
+ case ETHTOOL_GFLAGS:
+ rc = ethtool_get_value(dev, useraddr, ethcmd,
+ __ethtool_get_flags);
+ break;
+ case ETHTOOL_SFLAGS:
+ rc = ethtool_set_value(dev, useraddr, __ethtool_set_flags);
+ break;
+ case ETHTOOL_GRXFH:
+ case ETHTOOL_GRXRINGS:
+ case ETHTOOL_GRXCLSRLCNT:
+ case ETHTOOL_GRXCLSRULE:
+ case ETHTOOL_GRXCLSRLALL:
+ rc = ethtool_get_rxnfc(dev, ethcmd, useraddr);
+ break;
+ case ETHTOOL_SRXFH:
+ case ETHTOOL_SRXCLSRLDEL:
+ case ETHTOOL_SRXCLSRLINS:
+ rc = ethtool_set_rxnfc(dev, ethcmd, useraddr);
+ break;
+ case ETHTOOL_GCHANNELS:
+ rc = ethtool_get_channels(dev, useraddr);
+ break;
+ default:
+ rc = -EOPNOTSUPP;
+ }
+
+ if (dev->ethtool_ops->complete)
+ dev->ethtool_ops->complete(dev);
+
+ if (old_features != dev->features)
+ netdev_features_change(dev);
+
+ return rc;
+}
+
+int send_ethtool_ioctl(struct cmd_context *ctx, void *cmd)
+{
+ struct ifreq ifr = {0};
+ int ret;
+
+ strncpy(ifr.ifr_name, ctx->netdev, sizeof(ctx->netdev));
+ ifr.ifr_data = cmd;
+
+ rtnl_lock();
+ ret = dev_ethtool_kern(&init_net, &ifr);
+ rtnl_unlock();
+
+ return ret;
+}
+
+struct oecls_netdev_info *get_oecls_netdev_info(unsigned int index)
+{
+ if (index >= OECLS_MAX_NETDEV_NUM)
+ return NULL;
+ return &oecls_netdev_info_table[index];
+}
+
+static struct oecls_netdev_info *alloc_oecls_netdev_info(void)
+{
+ if (oecls_netdev_num >= OECLS_MAX_NETDEV_NUM)
+ return NULL;
+
+ return &oecls_netdev_info_table[oecls_netdev_num++];
+}
+
+static bool check_irq_name(const char *irq_name, struct oecls_netdev_info *oecls_dev)
+{
+ if (!strstr(irq_name, "TxRx") && !strstr(irq_name, "comp") && !strstr(irq_name, "rx"))
+ return false;
+
+ if (strstr(irq_name, oecls_dev->dev_name))
+ return true;
+
+ if (oecls_dev->netdev->dev.parent &&
+ strstr(irq_name, dev_name(oecls_dev->netdev->dev.parent)))
+ return true;
+
+ return false;
+}
+
+static void get_netdev_queue_info(struct oecls_netdev_info *oecls_dev)
+{
+ struct oecls_netdev_queue_info *rxq_info;
+ struct irq_desc *desc;
+ int irq;
+
+ for_each_irq_desc(irq, desc) {
+ if (!desc->action)
+ continue;
+ if (!desc->action->name)
+ continue;
+ if (!check_irq_name(desc->action->name, oecls_dev))
+ continue;
+
+ oecls_debug("irq=%d, [%s], rxq_id=%d\n", irq, desc->action->name,
+ oecls_dev->rxq_num);
+
+ if (oecls_dev->rxq_num >= OECLS_MAX_RXQ_NUM_PER_DEV)
+ break;
+ rxq_info = &oecls_dev->rxq[oecls_dev->rxq_num++];
+ rxq_info->irq = irq;
+ }
+}
+
+static int oecls_filter_enable(const char *dev_name, bool *old_state)
+{
+ struct ethtool_value eval = {0};
+ struct cmd_context ctx = {0};
+ int ret;
+
+ strncpy(ctx.netdev, dev_name, IFNAMSIZ);
+
+ eval.cmd = ETHTOOL_GFLAGS;
+ ret = send_ethtool_ioctl(&ctx, &eval);
+ if (ret != 0) {
+ oecls_error("get %s flags fail, ret:%d\n", dev_name, ret);
+ return ret;
+ }
+ if (eval.data & ETH_FLAG_NTUPLE) {
+ *old_state = true;
+ oecls_debug("%s ntuple is already on\n", dev_name);
+ return 0;
+ }
+
+ // Set ntuple feature
+ eval.cmd = ETHTOOL_SFLAGS;
+ eval.data |= ETH_FLAG_NTUPLE;
+ ret = send_ethtool_ioctl(&ctx, &eval);
+ if (ret != 0) {
+ oecls_error("set %s flags fail, ret:%d\n", dev_name, ret);
+ return ret;
+ }
+
+ // Get ntuple feature
+ eval.cmd = ETHTOOL_GFLAGS;
+ eval.data = 0;
+ ret = send_ethtool_ioctl(&ctx, &eval);
+ if (ret != 0) {
+ oecls_error("get %s flags fail, ret:%d\n", dev_name, ret);
+ return ret;
+ }
+ if (!(eval.data & ETH_FLAG_NTUPLE)) {
+ oecls_error("enable ntuple feature fail!\n");
+ return -EOPNOTSUPP;
+ }
+
+ return 0;
+}
+
+static void oecls_filter_restore(const char *dev_name, bool old_state)
+{
+ struct ethtool_value eval = {0};
+ struct cmd_context ctx = {0};
+ bool cur_filter_state;
+ int ret;
+
+ strncpy(ctx.netdev, dev_name, IFNAMSIZ);
+
+ eval.cmd = ETHTOOL_GFLAGS;
+ ret = send_ethtool_ioctl(&ctx, &eval);
+ if (ret != 0) {
+ oecls_error("get %s flags fail, ret:%d\n", dev_name, ret);
+ return;
+ }
+
+ cur_filter_state = (eval.data & ETH_FLAG_NTUPLE) ? true : false;
+ if (cur_filter_state == old_state)
+ return;
+
+ // Set ntuple feature
+ eval.cmd = ETHTOOL_SFLAGS;
+ if (old_state)
+ eval.data |= ETH_FLAG_NTUPLE;
+ else
+ eval.data &= ~ETH_FLAG_NTUPLE;
+ ret = send_ethtool_ioctl(&ctx, &eval);
+ if (ret != 0) {
+ oecls_error("set %s flags fail, ret:%d\n", dev_name, ret);
+ return;
+ }
+}
+
+static int init_single_oecls_dev(char *if_name, unsigned int length)
+{
+ struct oecls_netdev_info *oecls_dev;
+ char dev_name[IFNAMSIZ] = { 0 };
+ struct net_device *netdev;
+ int cpy_len = length < IFNAMSIZ ? length : IFNAMSIZ;
+ bool old_state = false;
+ int ret;
+
+ strncpy(dev_name, if_name, cpy_len);
+ netdev = dev_get_by_name(&init_net, dev_name);
+ if (!netdev) {
+ oecls_error("dev [%s] is not exist!\n", dev_name);
+ return -ENODEV;
+ }
+
+ if (!(netdev->flags & IFF_UP)) {
+ ret = -ENETDOWN;
+ oecls_error("dev:%s not up! flags=%d.\n", dev_name, netdev->flags);
+ goto out;
+ }
+
+ if (netdev->flags & IFF_LOOPBACK) {
+ ret = -EOPNOTSUPP;
+ oecls_error("Do not support loopback.\n");
+ goto out;
+ }
+
+ ret = oecls_filter_enable(dev_name, &old_state);
+ if (ret) {
+ oecls_error("dev [%s] not support ntuple! ret=%d\n", dev_name, ret);
+ goto out;
+ }
+
+ oecls_dev = alloc_oecls_netdev_info();
+ if (!oecls_dev) {
+ ret = -ENOMEM;
+ oecls_filter_restore(dev_name, old_state);
+ oecls_error("alloc oecls_dev fail! oecls_netdev_num:%d\n", oecls_netdev_num);
+ goto out;
+ }
+
+ memcpy_r(oecls_dev->dev_name, dev_name, IFNAMSIZ);
+ oecls_dev->old_filter_state = old_state;
+ oecls_dev->netdev = netdev;
+ get_netdev_queue_info(oecls_dev);
+ return 0;
+
+out:
+ dev_put(netdev);
+ return ret;
+}
+
+static void clean_oecls_netdev_info(void)
+{
+ struct oecls_netdev_info *oecls_dev;
+ struct net_device *netdev;
+ int devid;
+
+ for_each_oecls_netdev(devid, oecls_dev) {
+ oecls_filter_restore(oecls_dev->dev_name, oecls_dev->old_filter_state);
+ netdev = oecls_dev->netdev;
+ if (netdev) {
+ oecls_dev->netdev = NULL;
+ dev_put(netdev);
+ }
+ }
+
+ oecls_netdev_num = 0;
+}
+
+static int init_oecls_netdev_info(char *netdev_str)
+{
+ char *start = netdev_str;
+ char *end;
+ int err = -ENODEV;
+
+ while (*start != '\0') {
+ // skip start #
+ end = strchr(start, '#');
+ if (end == start) {
+ start++;
+ continue;
+ }
+
+ // find the last ifname
+ if (!end) {
+ err = init_single_oecls_dev(start, strlen(start));
+ break;
+ }
+
+ err = init_single_oecls_dev(start, end - start);
+ if (err)
+ break;
+ start = end + 1;
+ }
+
+ return err;
+}
+
+struct oecls_numa_info *get_oecls_numa_info(unsigned int nid)
+{
+ if (nid >= OECLS_MAX_NUMA_NUM)
+ return NULL;
+ return &oecls_numa_info_table[nid];
+}
+
+static void clean_oecls_numa_info(void)
+{
+ oecls_numa_num = 0;
+}
+
+static void init_numa_avail_cpus(int nid, struct oecls_numa_info *numa_info)
+{
+ int cpu;
+
+ oecls_debug("numa node %d: %*pb, %*pbl\n", nid, cpumask_pr_args(cpumask_of_node(nid)),
+ cpumask_pr_args(cpumask_of_node(nid)));
+
+ bitmap_zero(numa_info->avail_cpus, OECLS_MAX_CPU_NUM);
+ for_each_cpu(cpu, cpumask_of_node(nid)) {
+ if (cpu >= OECLS_MAX_CPU_NUM)
+ return;
+ set_bit(cpu, numa_info->avail_cpus);
+ }
+}
+
+static void init_numa_rxq_bitmap(int nid, struct oecls_numa_info *numa_info)
+{
+ struct oecls_numa_bound_dev_info *bound_dev;
+ struct oecls_netdev_info *oecls_dev;
+ int bound_rxq_num;
+ int rxq_id;
+ int devid;
+ int cpu;
+
+ for_each_oecls_netdev(devid, oecls_dev) {
+ bound_rxq_num = 0;
+ bound_dev = &numa_info->bound_dev[devid];
+ bitmap_zero(bound_dev->bitmap_rxq, OECLS_MAX_RXQ_NUM_PER_DEV);
+
+ for (rxq_id = 0; rxq_id < oecls_dev->rxq_num; rxq_id++) {
+ cpu = oecls_dev->rxq[rxq_id].affinity_cpu;
+ if (cpu_to_node(cpu) == nid) {
+ set_bit(rxq_id, bound_dev->bitmap_rxq);
+ bound_rxq_num++;
+ }
+ }
+
+ oecls_debug("nid:%d, dev_id:%d, dev:%s, rxq_num:%d, bit_num:%d, bitmap_rxq:%*pbl\n",
+ nid, devid, oecls_dev->dev_name, oecls_dev->rxq_num,
+ bound_rxq_num, OECLS_MAX_RXQ_NUM_PER_DEV, bound_dev->bitmap_rxq);
+ }
+}
+
+int alloc_rxq_id(int nid, int devid)
+{
+ struct oecls_numa_bound_dev_info *bound_dev;
+ struct oecls_numa_info *numa_info;
+ int rxq_id;
+
+ numa_info = get_oecls_numa_info(nid);
+ if (!numa_info) {
+ oecls_error("error nid:%d\n", nid);
+ return -EINVAL;
+ }
+
+ if (devid >= OECLS_MAX_NETDEV_NUM) {
+ oecls_error("error bound_dev index:%d\n", devid);
+ return -EINVAL;
+ }
+ bound_dev = &numa_info->bound_dev[devid];
+
+ rxq_id = find_first_bit(bound_dev->bitmap_rxq, OECLS_MAX_RXQ_NUM_PER_DEV);
+ if (rxq_id >= OECLS_MAX_RXQ_NUM_PER_DEV) {
+ oecls_error("error rxq_id:%d\n", rxq_id);
+ return -EINVAL;
+ }
+
+ clear_bit(rxq_id, bound_dev->bitmap_rxq);
+ oecls_debug("alloc nid:%d, dev_id:%d, rxq_id:%d\n", nid, devid, rxq_id);
+ return rxq_id;
+}
+
+void free_rxq_id(int nid, int devid, int rxq_id)
+{
+ struct oecls_numa_bound_dev_info *bound_dev;
+ struct oecls_numa_info *numa_info;
+
+ numa_info = get_oecls_numa_info(nid);
+ if (!numa_info) {
+ oecls_error("error nid:%d\n", nid);
+ return;
+ }
+
+ if (devid >= OECLS_MAX_NETDEV_NUM) {
+ oecls_error("error bound_dev index:%d\n", devid);
+ return;
+ }
+ bound_dev = &numa_info->bound_dev[devid];
+
+ if (rxq_id >= OECLS_MAX_RXQ_NUM_PER_DEV) {
+ oecls_error("error rxq_id:%d\n", rxq_id);
+ return;
+ }
+
+ if (test_bit(rxq_id, bound_dev->bitmap_rxq)) {
+ oecls_error("error nid:%d, devid:%d, rxq_id:%d\n", nid, devid, rxq_id);
+ return;
+ }
+
+ set_bit(rxq_id, bound_dev->bitmap_rxq);
+ oecls_debug("free nid:%d, dev_id:%d, rxq_id:%d\n", nid, devid, rxq_id);
+}
+
+static void init_oecls_numa_info(void)
+{
+ struct oecls_numa_info *numa_info;
+ unsigned int numa_num;
+ int nid;
+
+ numa_num = num_online_nodes();
+ if (numa_num > OECLS_MAX_NUMA_NUM) {
+ oecls_error("online numa num:%d is too much!\n", numa_num);
+ numa_num = OECLS_MAX_NUMA_NUM;
+ }
+ oecls_numa_num = numa_num;
+ oecls_debug("set oecls_numa_num=%d\n", numa_num);
+
+ for_each_oecls_numa(nid, numa_info)
+ init_numa_avail_cpus(nid, numa_info);
+}
+
+static int alloc_available_cpu(int nid, struct oecls_numa_info *numa_info)
+{
+ int cpu;
+
+ cpu = find_first_bit(numa_info->avail_cpus, OECLS_MAX_CPU_NUM);
+ if (cpu >= OECLS_MAX_CPU_NUM) {
+ oecls_error("no available cpus: nid=%d, cpu=%d\n", nid, cpu);
+ return -1;
+ }
+
+ clear_bit(cpu, numa_info->avail_cpus);
+ return cpu;
+}
+
+static void add_netdev_irq_affinity_cpu(struct oecls_netdev_info *oecls_dev, int rxq_id, int cpu)
+{
+ struct oecls_netdev_queue_info *rxq_info;
+
+ if (rxq_id >= OECLS_MAX_RXQ_NUM_PER_DEV)
+ return;
+
+ rxq_info = &oecls_dev->rxq[rxq_id];
+ rxq_info->affinity_cpu = cpu;
+}
+
+static void config_affinity_strategy_default(struct oecls_netdev_info *oecls_dev)
+{
+ struct oecls_numa_info *numa_info;
+ int rxq_num = oecls_dev->rxq_num;
+ int rxq_per_numa = rxq_num / oecls_numa_num;
+ int remain = rxq_num - rxq_per_numa * oecls_numa_num;
+ int numa_rxq_id;
+ int rxq_id;
+ int nid;
+ int cpu;
+
+ oecls_debug("dev=%s, rxq_num=%d, rxq_per_numa=%d, remain=%d\n", oecls_dev->dev_name,
+ rxq_num, rxq_per_numa, remain);
+
+ // average config rxq to every numa
+ for_each_oecls_numa(nid, numa_info) {
+ for (numa_rxq_id = 0; numa_rxq_id < rxq_per_numa; numa_rxq_id++) {
+ cpu = alloc_available_cpu(nid, numa_info);
+ if (cpu < 0)
+ break;
+
+ rxq_id = rxq_per_numa * nid + numa_rxq_id;
+ add_netdev_irq_affinity_cpu(oecls_dev, rxq_id, cpu);
+ }
+ }
+
+ if (!remain)
+ return;
+
+ // config remain rxq to every numa
+ numa_rxq_id = 0;
+ for_each_oecls_numa(nid, numa_info) {
+ if (numa_rxq_id >= remain)
+ break;
+ cpu = alloc_available_cpu(nid, numa_info);
+ if (cpu < 0)
+ break;
+
+ rxq_id = rxq_per_numa * oecls_numa_num + numa_rxq_id;
+ numa_rxq_id++;
+ add_netdev_irq_affinity_cpu(oecls_dev, rxq_id, cpu);
+ }
+}
+
+static void config_affinity_strategy_cluster(struct oecls_netdev_info *oecls_dev)
+{
+ int cluster_cpu_num = 8;
+ int cluster_num = num_online_cpus() / cluster_cpu_num;
+ int cluster_cpu_id = 0;
+ int rxq_id = 0;
+ int cluster;
+ int cpu;
+
+ // average config rxq to every cluster
+ while (rxq_id < oecls_dev->rxq_num) {
+ for (cluster = 0; cluster < cluster_num; cluster++) {
+ cpu = cluster * cluster_cpu_num + cluster_cpu_id;
+ if (rxq_id >= oecls_dev->rxq_num)
+ break;
+ add_netdev_irq_affinity_cpu(oecls_dev, rxq_id++, cpu);
+ }
+ cluster_cpu_id++;
+ }
+}
+
+static void config_affinity_strategy_16cores(struct oecls_netdev_info *oecls_dev)
+{
+ struct oecls_numa_info *numa_info;
+ int numa_start_cpu;
+ int numa_cpu_id;
+ int rxq_id = 0;
+ int nid;
+ int cpu;
+
+ // only use 16 cores of one numa
+ for_each_oecls_numa(nid, numa_info) {
+ numa_start_cpu = find_first_bit(numa_info->avail_cpus, OECLS_MAX_CPU_NUM);
+ for (numa_cpu_id = 0; numa_cpu_id < 16; numa_cpu_id++) {
+ cpu = numa_start_cpu + numa_cpu_id;
+
+ if (rxq_id >= oecls_dev->rxq_num)
+ break;
+ add_netdev_irq_affinity_cpu(oecls_dev, rxq_id++, cpu);
+ }
+ }
+}
+
+static void config_affinity_strategy(void)
+{
+ struct oecls_netdev_info *oecls_dev;
+ int devid;
+
+ for_each_oecls_netdev(devid, oecls_dev) {
+ switch (strategy) {
+ case 0:
+ config_affinity_strategy_default(oecls_dev);
+ break;
+ case 1:
+ config_affinity_strategy_cluster(oecls_dev);
+ break;
+ case 2:
+ config_affinity_strategy_16cores(oecls_dev);
+ break;
+ default:
+ config_affinity_strategy_default(oecls_dev);
+ break;
+ }
+ }
+}
+
+static inline void irq_set_affinity_wrapper(int rxq, int irq, int cpu)
+{
+ int err = 0;
+
+ err = irq_set_affinity(irq, get_cpu_mask(cpu));
+ oecls_debug("rxq=%d, irq=%d, cpu=%d, err=%d\n", rxq, irq, cpu, err);
+}
+
+static void enable_affinity_strategy(void)
+{
+ struct oecls_netdev_queue_info *rxq_info;
+ struct oecls_netdev_info *oecls_dev;
+ int rxq_id;
+ int devid;
+
+ for_each_oecls_netdev(devid, oecls_dev) {
+ for (rxq_id = 0; rxq_id < oecls_dev->rxq_num; rxq_id++) {
+ rxq_info = &oecls_dev->rxq[rxq_id];
+ irq_set_affinity_wrapper(rxq_id, rxq_info->irq, rxq_info->affinity_cpu);
+ }
+ }
+}
+
+static inline void netif_set_xps_queue_wrapper(struct net_device *netdev, int rxq_id,
+ const struct cpumask *cpu_mask)
+{
+ int err = 0;
+
+ err = netif_set_xps_queue(netdev, cpu_mask, rxq_id);
+ oecls_debug("name=%s, rxq_id=%d, mask=%*pbl, err=%d\n", netdev->name, rxq_id,
+ cpumask_pr_args(cpu_mask), err);
+}
+
+static void set_netdev_xps_queue(bool enable)
+{
+ const struct cpumask clear_mask = { 0 };
+ struct oecls_netdev_info *oecls_dev;
+ const struct cpumask *cpu_mask;
+ int rxq_id;
+ int devid;
+ int cpu;
+ int nid;
+
+ for_each_oecls_netdev(devid, oecls_dev) {
+ for (rxq_id = 0; rxq_id < oecls_dev->rxq_num; rxq_id++) {
+ cpu = oecls_dev->rxq[rxq_id].affinity_cpu;
+ nid = cpu_to_node(cpu);
+ if (enable)
+ cpu_mask = cpumask_of_node(nid);
+ else
+ cpu_mask = &clear_mask;
+
+ netif_set_xps_queue_wrapper(oecls_dev->netdev, rxq_id, cpu_mask);
+ }
+ }
+}
+
+static __init int oecls_init(void)
+{
+ struct oecls_numa_info *numa_info;
+ int nid;
+ int err;
+
+ oecls_debug("[init] mode=%d, ifname=[%s]\n", mode, ifname);
+ if (!check_params())
+ return -EINVAL;
+
+ init_oecls_l0_cache();
+ init_oecls_numa_info();
+ err = init_oecls_netdev_info(ifname);
+ if (err)
+ goto out;
+
+ // Set irq affinity
+ config_affinity_strategy();
+ enable_affinity_strategy();
+
+ // Calculate rxq bounded to one numa
+ for_each_oecls_numa(nid, numa_info)
+ init_numa_rxq_bitmap(nid, numa_info);
+
+#ifdef CONFIG_XPS
+ set_netdev_xps_queue(true);
+#endif
+
+ if (mode == 0)
+ oecls_ntuple_res_init();
+ else
+ oecls_flow_res_init();
+
+ return 0;
+out:
+ clean_oecls_netdev_info();
+ clean_oecls_numa_info();
+ clean_oecls_l0_cache();
+ return err;
+}
+
+static __exit void oecls_exit(void)
+{
+ oecls_debug("[exit] mode=%d\n", mode);
+ if (mode == 0)
+ oecls_ntuple_res_clean();
+ else
+ oecls_flow_res_clean();
+
+#ifdef CONFIG_XPS
+ set_netdev_xps_queue(false);
+#endif
+
+ clean_oecls_netdev_info();
+ clean_oecls_numa_info();
+ clean_oecls_l0_cache();
+}
+
+module_init(oecls_init);
+module_exit(oecls_exit);
+
+MODULE_DESCRIPTION("oenetcls");
+MODULE_LICENSE("GPL v2");
diff --git a/net/oenetcls/oenetcls_ntuple.c b/net/oenetcls/oenetcls_ntuple.c
new file mode 100644
index 000000000000..38d1f5df6ff1
--- /dev/null
+++ b/net/oenetcls/oenetcls_ntuple.c
@@ -0,0 +1,510 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/inetdevice.h>
+#include <linux/netdevice.h>
+#include <linux/rtnetlink.h>
+#include <linux/irq.h>
+#include <linux/irqdesc.h>
+#include <linux/inet.h>
+#include <linux/jhash.h>
+#include <net/sock.h>
+#include <trace/hooks/oenetcls.h>
+#include "oenetcls.h"
+
+struct oecls_sk_rule_list oecls_sk_rules;
+
+static void init_oecls_sk_rules(void)
+{
+ unsigned int i;
+
+ for (i = 0; i < OECLS_SK_RULE_HASHSIZE; i++)
+ INIT_HLIST_HEAD(oecls_sk_rules.hash + i);
+ mutex_init(&oecls_sk_rules.mutex);
+}
+
+static struct hlist_head *oecls_sk_rule_hash(u32 dip4, u16 dport)
+{
+ return oecls_sk_rules.hash + (jhash_2words(dip4, dport, 0) & OECLS_SK_RULE_HASHMASK);
+}
+
+static void add_sk_rule(int devid, u32 dip4, u16 dport, void *sk, int action,
+ int ruleid, int nid)
+{
+ struct hlist_head *hlist = oecls_sk_rule_hash(dip4, dport);
+ struct oecls_sk_rule *rule;
+
+ rule = alloc_from_l0(sizeof(struct oecls_sk_rule));
+ if (!rule)
+ return;
+ oecls_debug("alloc rule=%p\n", rule);
+
+ rule->sk = sk;
+ rule->dip4 = dip4;
+ rule->dport = dport;
+ rule->devid = devid;
+ rule->action = action;
+ rule->ruleid = ruleid;
+ rule->nid = nid;
+ hlist_add_head(&rule->node, hlist);
+}
+
+static void del_sk_rule(struct oecls_sk_rule *rule)
+{
+ hlist_del_init(&rule->node);
+ oecls_debug("del rule=%p\n", rule);
+ free_to_l0(rule);
+}
+
+static struct oecls_sk_rule *get_sk_rule(int devid, u32 dip4, u16 dport)
+{
+ struct hlist_head *hlist = oecls_sk_rule_hash(dip4, dport);
+ struct oecls_sk_rule *rule = NULL;
+
+ hlist_for_each_entry(rule, hlist, node) {
+ if (rule->devid == devid && rule->dip4 == dip4 && rule->dport == dport)
+ break;
+ }
+ return rule;
+}
+
+static bool reuseport_check(int devid, u32 dip4, u16 dport)
+{
+ return !!get_sk_rule(devid, dip4, dport);
+}
+
+static u32 get_first_ip4_addr(struct net *net)
+{
+ struct in_device *in_dev;
+ struct net_device *dev;
+ struct in_ifaddr *ifa;
+ u32 dip4 = 0;
+
+ rtnl_lock();
+ rcu_read_lock();
+ for_each_netdev(net, dev) {
+ if (dev->flags & IFF_LOOPBACK || !(dev->flags & IFF_UP))
+ continue;
+ in_dev = __in_dev_get_rcu(dev);
+ if (!in_dev)
+ continue;
+
+ in_dev_for_each_ifa_rcu(ifa, in_dev) {
+ if (!strcmp(dev->name, ifa->ifa_label)) {
+ dip4 = ifa->ifa_local;
+ oecls_debug("dev: %s, dip4: 0x%x\n", dev->name, dip4);
+ goto out;
+ }
+ }
+ }
+out:
+ rcu_read_unlock();
+ rtnl_unlock();
+ return dip4;
+}
+
+static void get_sk_rule_addr(struct sock *sk, u32 *dip4, u16 *dport)
+{
+ *dport = htons(sk->sk_num);
+
+ if (!match_ip_flag) {
+ *dip4 = 0;
+ return;
+ }
+
+ if (sk->sk_rcv_saddr)
+ *dip4 = sk->sk_rcv_saddr;
+ else
+ *dip4 = get_first_ip4_addr(sock_net(sk));
+}
+
+static int rxclass_rule_del(struct cmd_context *ctx, __u32 loc)
+{
+ struct ethtool_rxnfc nfccmd;
+ int err;
+
+ nfccmd.cmd = ETHTOOL_SRXCLSRLDEL;
+ nfccmd.fs.location = loc;
+ err = send_ethtool_ioctl(ctx, &nfccmd);
+ if (err < 0)
+ oecls_debug("rmgr: Cannot delete RX class rule, loc:%u\n", loc);
+ return err;
+}
+
+static int rmgr_ins(struct rmgr_ctrl *rmgr, __u32 loc)
+{
+ if (loc >= rmgr->size) {
+ oecls_error("rmgr: Location out of range\n");
+ return -1;
+ }
+
+ set_bit(loc, rmgr->slot);
+ return 0;
+}
+
+static int rmgr_find_empty_slot(struct rmgr_ctrl *rmgr, struct ethtool_rx_flow_spec *fsp)
+{
+ __u32 loc, slot_num;
+
+ if (rmgr->driver_select)
+ return 0;
+
+ loc = rmgr->size - 1;
+ slot_num = loc / BITS_PER_LONG;
+ if (!~(rmgr->slot[slot_num] | (~1UL << rmgr->size % BITS_PER_LONG))) {
+ loc -= 1 + (loc % BITS_PER_LONG);
+ slot_num--;
+ }
+
+ while (loc < rmgr->size && !~(rmgr->slot[slot_num])) {
+ loc -= BITS_PER_LONG;
+ slot_num--;
+ }
+
+ while (loc < rmgr->size && test_bit(loc, rmgr->slot))
+ loc--;
+
+ if (loc < rmgr->size) {
+ fsp->location = loc;
+ return rmgr_ins(rmgr, loc);
+ }
+
+ return -1;
+}
+
+static int rxclass_get_dev_info(struct cmd_context *ctx, __u32 *count, int *driver_select)
+{
+ struct ethtool_rxnfc nfccmd;
+ int err;
+
+ nfccmd.cmd = ETHTOOL_GRXCLSRLCNT;
+ nfccmd.data = 0;
+ err = send_ethtool_ioctl(ctx, &nfccmd);
+ *count = nfccmd.rule_cnt;
+ if (driver_select)
+ *driver_select = !!(nfccmd.data & RX_CLS_LOC_SPECIAL);
+ if (err < 0)
+ oecls_debug("rxclass: Cannot get RX class rule count\n");
+
+ return err;
+}
+
+static int rmgr_init(struct cmd_context *ctx, struct rmgr_ctrl *rmgr)
+{
+ struct ethtool_rxnfc *nfccmd;
+ __u32 *rule_locs;
+ int i, err = 0;
+
+ memset(rmgr, 0, sizeof(*rmgr));
+ err = rxclass_get_dev_info(ctx, &rmgr->n_rules, &rmgr->driver_select);
+ if (err < 0)
+ return err;
+
+ if (rmgr->driver_select)
+ return err;
+
+ nfccmd = kzalloc(sizeof(*nfccmd) + (rmgr->n_rules * sizeof(__u32)), GFP_ATOMIC);
+ if (!nfccmd) {
+ oecls_error("rmgr: Cannot allocate memory for RX class rule locations\n");
+ err = -ENOMEM;
+ goto out;
+ }
+
+ nfccmd->cmd = ETHTOOL_GRXCLSRLALL;
+ nfccmd->rule_cnt = rmgr->n_rules;
+ err = send_ethtool_ioctl(ctx, nfccmd);
+ if (err < 0) {
+ oecls_debug("rmgr: Cannot get RX class rules\n");
+ goto out;
+ }
+
+ rmgr->size = nfccmd->data;
+ if (rmgr->size == 0 || rmgr->size < rmgr->n_rules) {
+ oecls_error("rmgr: Invalid RX class rules table size\n");
+ err = -EINVAL;
+ goto out;
+ }
+
+ rmgr->slot = kzalloc(BITS_TO_LONGS(rmgr->size) * sizeof(long), GFP_ATOMIC);
+ if (!rmgr->slot) {
+ oecls_error("rmgr: Cannot allocate memory for RX class rules\n");
+ err = -ENOMEM;
+ goto out;
+ }
+
+ rule_locs = nfccmd->rule_locs;
+ for (i = 0; i < rmgr->n_rules; i++) {
+ err = rmgr_ins(rmgr, rule_locs[i]);
+ if (err < 0)
+ break;
+ }
+
+out:
+ kfree(nfccmd);
+ return err;
+}
+
+static void rmgr_cleanup(struct rmgr_ctrl *rmgr)
+{
+ kfree(rmgr->slot);
+ rmgr->slot = NULL;
+ rmgr->size = 0;
+}
+
+static int rmgr_set_location(struct cmd_context *ctx,
+ struct ethtool_rx_flow_spec *fsp)
+{
+ struct rmgr_ctrl rmgr;
+ int ret;
+
+ ret = rmgr_init(ctx, &rmgr);
+ if (ret < 0)
+ goto out;
+
+ ret = rmgr_find_empty_slot(&rmgr, fsp);
+out:
+ rmgr_cleanup(&rmgr);
+ return ret;
+}
+
+static int rxclass_rule_ins(struct cmd_context *ctx,
+ struct ethtool_rx_flow_spec *fsp, u32 rss_context)
+{
+ struct ethtool_rxnfc nfccmd;
+ u32 loc = fsp->location;
+ int ret;
+
+ if (loc & RX_CLS_LOC_SPECIAL) {
+ ret = rmgr_set_location(ctx, fsp);
+ if (ret < 0)
+ return ret;
+ }
+
+ nfccmd.cmd = ETHTOOL_SRXCLSRLINS;
+ nfccmd.rss_context = rss_context;
+ nfccmd.fs = *fsp;
+ ret = send_ethtool_ioctl(ctx, &nfccmd);
+ if (ret < 0) {
+ oecls_debug("Can not insert the clasification rule\n");
+ return ret;
+ }
+
+ if (loc & RX_CLS_LOC_SPECIAL)
+ oecls_debug("Added rule with ID %d\n", nfccmd.fs.location);
+
+ return 0;
+}
+
+static void flow_spec_to_ntuple(struct ethtool_rx_flow_spec *fsp,
+ struct ethtool_rx_ntuple_flow_spec *ntuple)
+{
+ int i;
+
+ memset(ntuple, ~0, sizeof(*ntuple));
+ ntuple->flow_type = fsp->flow_type;
+ ntuple->action = fsp->ring_cookie;
+ memcpy_r(&ntuple->h_u, &fsp->h_u, sizeof(fsp->h_u));
+ memcpy_r(&ntuple->m_u, &fsp->m_u, sizeof(fsp->m_u));
+ for (i = 0; i < sizeof(ntuple->m_u); i++)
+ ntuple->m_u.hdata[i] ^= 0xFF;
+ ntuple->flow_type &= ~FLOW_EXT;
+}
+
+static int do_srxntuple(struct cmd_context *ctx, struct ethtool_rx_flow_spec *fsp)
+{
+ struct ethtool_rx_ntuple ntuplecmd;
+ struct ethtool_value eval;
+ int ret = 0;
+
+ flow_spec_to_ntuple(fsp, &ntuplecmd.fs);
+
+ eval.cmd = ETHTOOL_GFLAGS;
+ ret = send_ethtool_ioctl(ctx, &eval);
+ if (ret || !(eval.data & ETH_FLAG_NTUPLE))
+ return -1;
+
+ ntuplecmd.cmd = ETHTOOL_SRXNTUPLE;
+ ret = send_ethtool_ioctl(ctx, &ntuplecmd);
+ if (ret)
+ oecls_debug("Cannot add new rule via N-tuple, ret:%d\n", ret);
+
+ return ret;
+}
+
+static int cfg_ethtool_rule(struct cmd_context *ctx, bool is_del)
+{
+ struct ethtool_rx_flow_spec *fsp, rx_rule_fs;
+ u32 rss_context = 0;
+ int ret;
+
+ oecls_debug("is_del:%d netdev:%s, dip4:%pI4, dport:%d, action:%d, ruleid:%u, del_ruleid:%u\n",
+ is_del, ctx->netdev, &ctx->dip4, ntohs(ctx->dport), ctx->action, ctx->ruleid,
+ ctx->del_ruleid);
+
+ if (is_del)
+ return rxclass_rule_del(ctx, ctx->del_ruleid);
+
+ ctx->ret_loc = -1;
+
+ fsp = &rx_rule_fs;
+ memset(fsp, 0, sizeof(*fsp));
+ fsp->flow_type = TCP_V4_FLOW;
+ fsp->location = RX_CLS_LOC_ANY;
+ fsp->h_u.tcp_ip4_spec.ip4dst = ctx->dip4;
+ fsp->h_u.tcp_ip4_spec.pdst = ctx->dport;
+ if (ctx->dip4)
+ fsp->m_u.tcp_ip4_spec.ip4dst = (u32)~0ULL;
+ fsp->m_u.tcp_ip4_spec.pdst = (u16)~0ULL;
+ if (ctx->ruleid)
+ fsp->location = ctx->ruleid;
+ fsp->ring_cookie = ctx->action;
+
+ ret = do_srxntuple(ctx, &rx_rule_fs);
+ if (!ret)
+ return 0;
+
+ ret = rxclass_rule_ins(ctx, &rx_rule_fs, rss_context);
+ if (!ret)
+ ctx->ret_loc = rx_rule_fs.location;
+ return ret;
+}
+
+static void del_ntuple_rule(struct sock *sk)
+{
+ struct oecls_netdev_info *oecls_dev;
+ struct cmd_context ctx = { 0 };
+ struct oecls_sk_rule *rule;
+ int devid;
+ u16 dport;
+ u32 dip4;
+ int err;
+
+ get_sk_rule_addr(sk, &dip4, &dport);
+
+ mutex_lock(&oecls_sk_rules.mutex);
+ for_each_oecls_netdev(devid, oecls_dev) {
+ strncpy(ctx.netdev, oecls_dev->dev_name, IFNAMSIZ);
+ rule = get_sk_rule(devid, dip4, dport);
+ if (!rule) {
+ oecls_debug("rule not found! sk:%p, devid:%d, dip4:0x%x, dport:%d\n", sk,
+ devid, dip4, dport);
+ continue;
+ }
+
+ // Config Ntuple rule to dev
+ ctx.del_ruleid = rule->ruleid;
+ err = cfg_ethtool_rule(&ctx, true);
+ if (err) {
+ oecls_error("del sk:%p, nid:%d, devid:%d, action:%d, ruleid:%d, err:%d\n",
+ sk, rule->nid, devid, rule->action, rule->ruleid, err);
+ }
+
+ // Free the bound queue
+ free_rxq_id(rule->nid, devid, rule->action);
+
+ // Delete sk rule
+ del_sk_rule(rule);
+ }
+ mutex_unlock(&oecls_sk_rules.mutex);
+}
+
+static void add_ntuple_rule(struct sock *sk)
+{
+ struct oecls_netdev_info *oecls_dev;
+ struct cmd_context ctx = { 0 };
+ int cpu = smp_processor_id();
+ int nid = cpu_to_node(cpu);
+ int rxq_id;
+ int devid;
+ int err;
+
+ if (check_appname(current->comm))
+ return;
+ get_sk_rule_addr(sk, &ctx.dip4, &ctx.dport);
+
+ mutex_lock(&oecls_sk_rules.mutex);
+ for_each_oecls_netdev(devid, oecls_dev) {
+ strncpy(ctx.netdev, oecls_dev->dev_name, IFNAMSIZ);
+ if (reuseport_check(devid, ctx.dip4, ctx.dport)) {
+ oecls_error("dip4:0x%x, dport:%d reuse!\n", ctx.dip4, ctx.dport);
+ continue;
+ }
+
+ // Calculate the bound queue
+ rxq_id = alloc_rxq_id(nid, devid);
+ if (rxq_id < 0)
+ continue;
+
+ // Config Ntuple rule to dev
+ ctx.action = (u16)rxq_id;
+ err = cfg_ethtool_rule(&ctx, false);
+ if (err) {
+ oecls_error("add sk:%p, nid:%d, devid:%d, action:%d, ruleid:%d, err:%d\n",
+ sk, nid, devid, ctx.action, ctx.ret_loc, err);
+ continue;
+ }
+
+ // Add sk rule
+ add_sk_rule(devid, ctx.dip4, ctx.dport, sk, ctx.action, ctx.ret_loc, nid);
+ }
+ mutex_unlock(&oecls_sk_rules.mutex);
+}
+
+static void ethtool_cfg_rxcls(void *data, struct sock *sk, int is_del)
+{
+ if (sk->sk_state != TCP_LISTEN)
+ return;
+
+ if (sk->sk_family != AF_INET && sk->sk_family != AF_INET6)
+ return;
+
+ oecls_debug("[cpu:%d] app:%s, sk:%p, is_del:%d, ip:0x%x, port:0x%x\n", smp_processor_id(),
+ current->comm, sk, is_del, sk->sk_rcv_saddr, sk->sk_num);
+
+ if (is_del)
+ del_ntuple_rule(sk);
+ else
+ add_ntuple_rule(sk);
+}
+
+static void clean_oecls_sk_rules(void)
+{
+ struct oecls_netdev_info *oecls_dev;
+ struct cmd_context ctx = { 0 };
+ struct oecls_sk_rule *rule;
+ struct hlist_head *hlist;
+ struct hlist_node *n;
+ unsigned int i;
+ int err;
+
+ mutex_lock(&oecls_sk_rules.mutex);
+ for (i = 0; i < OECLS_SK_RULE_HASHSIZE; i++) {
+ hlist = &oecls_sk_rules.hash[i];
+
+ hlist_for_each_entry_safe(rule, n, hlist, node) {
+ oecls_dev = get_oecls_netdev_info(rule->devid);
+ if (!oecls_dev)
+ continue;
+ strncpy(ctx.netdev, oecls_dev->dev_name, IFNAMSIZ);
+ ctx.del_ruleid = rule->ruleid;
+ err = cfg_ethtool_rule(&ctx, true);
+ oecls_debug("sk:%p, dev_id:%d, action:%d, ruleid:%d, err:%d\n", rule->sk,
+ rule->devid, rule->action, rule->ruleid, err);
+
+ hlist_del(&rule->node);
+ oecls_debug("clean rule=%p\n", rule);
+ free_to_l0(rule);
+ }
+ }
+ mutex_unlock(&oecls_sk_rules.mutex);
+}
+
+void oecls_ntuple_res_init(void)
+{
+ init_oecls_sk_rules();
+ register_trace_ethtool_cfg_rxcls(ðtool_cfg_rxcls, NULL);
+}
+
+void oecls_ntuple_res_clean(void)
+{
+ unregister_trace_ethtool_cfg_rxcls(ðtool_cfg_rxcls, NULL);
+ clean_oecls_sk_rules();
+}
--
2.34.1
2
1
[openeuler:OLK-5.10 2938/2938] drivers/irqchip/irq-gic-v3-its.c:504:6: warning: no previous prototype for 'build_devid_pools'
by kernel test robot 30 May '25
by kernel test robot 30 May '25
30 May '25
tree: https://gitee.com/openeuler/kernel.git OLK-5.10
head: b5ec139da1a4916dfc9644ee3d9426dbd6f23c13
commit: af552f9d957def499065ed6e194bcd5f0c75dbc5 [2938/2938] irqchip/gic-v3-its: Move build_devid_pools from its to acpi iort init
config: arm64-allnoconfig (https://download.01.org/0day-ci/archive/20250530/202505301434.xq8uzhGR-lkp@…)
compiler: aarch64-linux-gcc (GCC) 15.1.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250530/202505301434.xq8uzhGR-lkp@…)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp(a)intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202505301434.xq8uzhGR-lkp@intel.com/
All warnings (new ones prefixed by >>):
>> drivers/irqchip/irq-gic-v3-its.c:504:6: warning: no previous prototype for 'build_devid_pools' [-Wmissing-prototypes]
504 | void build_devid_pools(void)
| ^~~~~~~~~~~~~~~~~
vim +/build_devid_pools +504 drivers/irqchip/irq-gic-v3-its.c
500
501 /*
502 * Currently we only build *one* devid pool.
503 */
> 504 void build_devid_pools(void)
505 {
506 struct its_node *its;
507
508 its = list_first_entry(&its_nodes, struct its_node, entry);
509 if (readl_relaxed(its->base + GITS_IIDR) != 0x00051736)
510 return;
511
512 if (!probe_devid_pool_one())
513 rsv_devid_pool_cap = true;
514
515 if (rsv_devid_pool_cap)
516 pr_info("ITS: reserved device id pools enabled\n");
517 }
518
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
1
0
[openeuler:OLK-6.6 2339/2339] kernel/sched/core.c:158:5: warning: no previous prototype for function 'sched_task_is_throttled'
by kernel test robot 30 May '25
by kernel test robot 30 May '25
30 May '25
tree: https://gitee.com/openeuler/kernel.git OLK-6.6
head: f528a77f54d3c07d0b7feb879d2ba15275a4b3ff
commit: b1415ae822e0f2c88266c3f6f8e2b962baa9a579 [2339/2339] interference: Exclude throttle time from rundelay
config: x86_64-buildonly-randconfig-005-20250530 (https://download.01.org/0day-ci/archive/20250530/202505301346.VRD8YS7H-lkp@…)
compiler: clang version 20.1.2 (https://github.com/llvm/llvm-project 58df0ef89dd64126512e4ee27b4ac3fd8ddf6247)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250530/202505301346.VRD8YS7H-lkp@…)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp(a)intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202505301346.VRD8YS7H-lkp@intel.com/
All warnings (new ones prefixed by >>):
>> kernel/sched/core.c:158:5: warning: no previous prototype for function 'sched_task_is_throttled' [-Wmissing-prototypes]
158 | int sched_task_is_throttled(struct task_struct *p, int cpu)
| ^
kernel/sched/core.c:158:1: note: declare 'static' if the function is not intended to be used outside of this translation unit
158 | int sched_task_is_throttled(struct task_struct *p, int cpu)
| ^
| static
kernel/sched/core.c:6465:1: warning: unused function 'class_core_lock_lock_ptr' [-Wunused-function]
6465 | DEFINE_LOCK_GUARD_1(core_lock, int,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
6466 | sched_core_lock(*_T->lock, &_T->flags),
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
6467 | sched_core_unlock(*_T->lock, &_T->flags),
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
6468 | unsigned long flags)
| ~~~~~~~~~~~~~~~~~~~~
include/linux/cleanup.h:229:49: note: expanded from macro 'DEFINE_LOCK_GUARD_1'
229 | __DEFINE_CLASS_IS_CONDITIONAL(_name, false); \
| ^
230 | __DEFINE_UNLOCK_GUARD(_name, _type, _unlock, __VA_ARGS__) \
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
include/linux/cleanup.h:205:21: note: expanded from macro '\
__DEFINE_UNLOCK_GUARD'
205 | static inline void *class_##_name##_lock_ptr(class_##_name##_t *_T) \
| ^~~~~~~~~~~~~~~~~~~~~~~~
<scratch space>:236:1: note: expanded from here
236 | class_core_lock_lock_ptr
| ^~~~~~~~~~~~~~~~~~~~~~~~
2 warnings generated.
vim +/sched_task_is_throttled +158 kernel/sched/core.c
156
157 #if defined(CONFIG_SCHED_CORE) || defined(CONFIG_CGROUP_IFS)
> 158 int sched_task_is_throttled(struct task_struct *p, int cpu)
159 {
160 if (p->sched_class->task_is_throttled)
161 return p->sched_class->task_is_throttled(p, cpu);
162
163 return 0;
164 }
165 #endif
166
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
1
0
Jinqian Yang (1):
KVM: arm64: fix memory leak in TLBI
arch/arm64/kvm/hisilicon/hisi_virt.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
--
2.33.0
2
2
Jinqian Yang (1):
KVM: arm64: fix memory leak in TLBI
arch/arm64/kvm/hisilicon/hisi_virt.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
--
2.33.0
2
2
[PATCH OLK-6.6 0/3] LoongArch: Fix bugs in early_numa_add_cpu/huge_pte_offset/__alloc_pages_slowpath
by Hongchen Zhang 30 May '25
by Hongchen Zhang 30 May '25
30 May '25
Huacai Chen (1):
LoongArch: Fix early_numa_add_cpu() usage for FDT systems
Tianyang Zhang (2):
LoongArch: Fix panic caused by NULL pmd in huge_pte_offset()
mm/page_alloc.c: avoid infinite retries caused by cpuset race
arch/loongarch/kernel/smp.c | 2 +-
arch/loongarch/mm/hugetlbpage.c | 3 ++-
mm/page_alloc.c | 8 ++++++++
3 files changed, 11 insertions(+), 2 deletions(-)
--
2.33.0
2
4
From: Frederic Weisbecker <frederic(a)kernel.org>
mainline inclusion
from mainline-v6.15-rc2
commit 56799bc035658738f362acec3e7647bb84e68933
category: bugfix
bugzilla: https://gitee.com/src-openeuler/kernel/issues/IC5BL4
CVE: CVE-2025-37747
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id…
--------------------------------
Perf can hang while freeing a sigtrap event if a related deferred
signal hadn't managed to be sent before the file got closed:
perf_event_overflow()
task_work_add(perf_pending_task)
fput()
task_work_add(____fput())
task_work_run()
____fput()
perf_release()
perf_event_release_kernel()
_free_event()
perf_pending_task_sync()
task_work_cancel() -> FAILED
rcuwait_wait_event()
Once task_work_run() is running, the list of pending callbacks is
removed from the task_struct and from this point on task_work_cancel()
can't remove any pending and not yet started work items, hence the
task_work_cancel() failure and the hang on rcuwait_wait_event().
Task work could be changed to remove one work at a time, so a work
running on the current task can always cancel a pending one, however
the wait / wake design is still subject to inverted dependencies when
remote targets are involved, as pictured by Oleg:
T1 T2
fd = perf_event_open(pid => T2->pid); fd = perf_event_open(pid => T1->pid);
close(fd) close(fd)
<IRQ> <IRQ>
perf_event_overflow() perf_event_overflow()
task_work_add(perf_pending_task) task_work_add(perf_pending_task)
</IRQ> </IRQ>
fput() fput()
task_work_add(____fput()) task_work_add(____fput())
task_work_run() task_work_run()
____fput() ____fput()
perf_release() perf_release()
perf_event_release_kernel() perf_event_release_kernel()
_free_event() _free_event()
perf_pending_task_sync() perf_pending_task_sync()
rcuwait_wait_event() rcuwait_wait_event()
Therefore the only option left is to acquire the event reference count
upon queueing the perf task work and release it from the task work, just
like it was done before 3a5465418f5f ("perf: Fix event leak upon exec and file release")
but without the leaks it fixed.
Some adjustments are necessary to make it work:
* A child event might dereference its parent upon freeing. Care must be
taken to release the parent last.
* Some places assuming the event doesn't have any reference held and
therefore can be freed right away must instead put the reference and
let the reference counting to its job.
Reported-by: "Yi Lai" <yi1.lai(a)linux.intel.com>
Closes: https://lore.kernel.org/all/Zx9Losv4YcJowaP%2F@ly-workstation/
Reported-by: syzbot+3c4321e10eea460eb606(a)syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/673adf75.050a0220.87769.0024.GAE@google.com/
Fixes: 3a5465418f5f ("perf: Fix event leak upon exec and file release")
Signed-off-by: Frederic Weisbecker <frederic(a)kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz(a)infradead.org>
Link: https://lkml.kernel.org/r/20250304135446.18905-1-frederic@kernel.org
Conflicts:
kernel/events/core.c
include/linux/perf_event.h
[To apply this mainline patch, first revert the 008ebcfec23d ("[Huawei]
perf: Fix kabi broken of struct perf_event"). Then pick mainline patch
16b9569df9d2 ("perf: Don't disable preemption in perf_pending_task().").
Finally, pick this fix patch and resolve conflicts and merged all patch
together.]
Signed-off-by: Luo Gengkun <luogengkun2(a)huawei.com>
---
include/linux/perf_event.h | 2 +-
kernel/events/core.c | 65 +++++++++++---------------------------
2 files changed, 19 insertions(+), 48 deletions(-)
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 420f15ddce39..826fb16906fe 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -851,7 +851,7 @@ struct perf_event {
*/
__u32 orig_type;
- KABI_USE(1, struct rcuwait pending_work_wait)
+ KABI_RESERVE(1)
KABI_RESERVE(2)
KABI_RESERVE(3)
KABI_RESERVE(4)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 69cdf8aabf52..f042d6101932 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2294,6 +2294,7 @@ event_sched_out(struct perf_event *event, struct perf_event_context *ctx)
!event->pending_work &&
!task_work_add(current, &event->pending_task, TWA_RESUME)) {
event->pending_work = 1;
+ WARN_ON_ONCE(!atomic_long_inc_not_zero(&event->refcount));
} else {
local_dec(&event->ctx->nr_pending);
}
@@ -5187,35 +5188,9 @@ static bool exclusive_event_installable(struct perf_event *event,
static void perf_addr_filters_splice(struct perf_event *event,
struct list_head *head);
-static void perf_pending_task_sync(struct perf_event *event)
-{
- struct callback_head *head = &event->pending_task;
-
- if (!event->pending_work)
- return;
- /*
- * If the task is queued to the current task's queue, we
- * obviously can't wait for it to complete. Simply cancel it.
- */
- if (task_work_cancel(current, head)) {
- event->pending_work = 0;
- local_dec(&event->ctx->nr_pending);
- return;
- }
-
- /*
- * All accesses related to the event are within the same
- * non-preemptible section in perf_pending_task(). The RCU
- * grace period before the event is freed will make sure all
- * those accesses are complete by then.
- */
- rcuwait_wait_event(&event->pending_work_wait, !event->pending_work, TASK_UNINTERRUPTIBLE);
-}
-
static void _free_event(struct perf_event *event)
{
irq_work_sync(&event->pending_irq);
- perf_pending_task_sync(event);
unaccount_event(event);
@@ -5340,10 +5315,17 @@ static void perf_remove_from_owner(struct perf_event *event)
static void put_event(struct perf_event *event)
{
+ struct perf_event *parent;
+
if (!atomic_long_dec_and_test(&event->refcount))
return;
+ parent = event->parent;
_free_event(event);
+
+ /* Matches the refcount bump in inherit_event() */
+ if (parent)
+ put_event(parent);
}
/*
@@ -5427,11 +5409,6 @@ int perf_event_release_kernel(struct perf_event *event)
if (tmp == child) {
perf_remove_from_context(child, DETACH_GROUP);
list_move(&child->child_list, &free_list);
- /*
- * This matches the refcount bump in inherit_event();
- * this can't be the last reference.
- */
- put_event(event);
} else {
var = &ctx->refcount;
}
@@ -5457,7 +5434,8 @@ int perf_event_release_kernel(struct perf_event *event)
void *var = &child->ctx->refcount;
list_del(&child->child_list);
- free_event(child);
+ /* Last reference unless ->pending_task work is pending */
+ put_event(child);
/*
* Wake any perf_event_free_task() waiting for this event to be
@@ -5468,7 +5446,11 @@ int perf_event_release_kernel(struct perf_event *event)
}
no_ctx:
- put_event(event); /* Must be the 'last' reference */
+ /*
+ * Last reference unless ->pending_task work is pending on this event
+ * or any of its children.
+ */
+ put_event(event);
return 0;
}
EXPORT_SYMBOL_GPL(perf_event_release_kernel);
@@ -6857,12 +6839,6 @@ static void perf_pending_task(struct callback_head *head)
struct perf_event *event = container_of(head, struct perf_event, pending_task);
int rctx;
- /*
- * All accesses to the event must belong to the same implicit RCU read-side
- * critical section as the ->pending_work reset. See comment in
- * perf_pending_task_sync().
- */
- preempt_disable_notrace();
/*
* If we 'fail' here, that's OK, it means recursion is already disabled
* and we won't recurse 'further'.
@@ -6873,12 +6849,11 @@ static void perf_pending_task(struct callback_head *head)
event->pending_work = 0;
perf_sigtrap(event);
local_dec(&event->ctx->nr_pending);
- rcuwait_wake_up(&event->pending_work_wait);
}
+ put_event(event);
if (rctx >= 0)
perf_swevent_put_recursion_context(rctx);
- preempt_enable_notrace();
}
#ifdef CONFIG_GUEST_PERF_EVENTS
@@ -11990,7 +11965,6 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
init_waitqueue_head(&event->waitq);
init_irq_work(&event->pending_irq, perf_pending_irq);
init_task_work(&event->pending_task, perf_pending_task);
- rcuwait_init(&event->pending_work_wait);
mutex_init(&event->mmap_mutex);
raw_spin_lock_init(&event->addr_filters.lock);
@@ -13143,8 +13117,7 @@ perf_event_exit_event(struct perf_event *event, struct perf_event_context *ctx)
* Kick perf_poll() for is_event_hup();
*/
perf_event_wakeup(parent_event);
- free_event(event);
- put_event(parent_event);
+ put_event(event);
return;
}
@@ -13262,13 +13235,11 @@ static void perf_free_event(struct perf_event *event,
list_del_init(&event->child_list);
mutex_unlock(&parent->child_mutex);
- put_event(parent);
-
raw_spin_lock_irq(&ctx->lock);
perf_group_detach(event);
list_del_event(event, ctx);
raw_spin_unlock_irq(&ctx->lock);
- free_event(event);
+ put_event(event);
}
/*
--
2.34.1
2
1