[PATCH AUTOSEL 6.15 03/23] um: vector: Reduce stack usage in vector_eth_configure()

[PATCH AUTOSEL 6.15 03/23] um: vector: Reduce stack usage in vector_eth_configure()

[Sasha Levin]

From: Tiwei Bie <tiwei.btw@antgroup.com>

[ Upstream commit 2d65fc13be85c336c56af7077f08ccd3a3a15a4a ]

When compiling with clang (19.1.7), initializing *vp using a compound
literal may result in excessive stack usage. Fix it by initializing the
required fields of *vp individually.

Without this patch:

$ objdump -d arch/um/drivers/vector_kern.o | ./scripts/checkstack.pl x86_64 0
...
0x0000000000000540 vector_eth_configure [vector_kern.o]:1472
...

With this patch:

$ objdump -d arch/um/drivers/vector_kern.o | ./scripts/checkstack.pl x86_64 0
...
0x0000000000000540 vector_eth_configure [vector_kern.o]:208
...

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202506221017.WtB7Usua-lkp@intel.com/
Signed-off-by: Tiwei Bie <tiwei.btw@antgroup.com>
Link: https://patch.msgid.link/20250623110829.314864-1-tiwei.btw@antgroup.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---

**YES**

This commit should be backported to stable kernel trees for the
following reasons:

1. **Fixes a real bug**: The commit addresses excessive stack usage
   (1472 bytes) that can lead to stack overflow, especially problematic
   on systems with limited kernel stack space. This is a legitimate bug
   that affects system stability.

2. **Compiler-specific issue with real impact**: While triggered by
   clang 19.1.7's handling of compound literals, the resulting stack
   usage of 1472 bytes is genuinely excessive and dangerous regardless
   of the compiler quirk that exposed it.

3. **Simple and safe fix**: The change is purely mechanical - converting
   from compound literal initialization to field-by-field
   initialization:
  ```c
  // From:
  *vp = ((struct vector_private) { .field = value, ... });
  // To:
  vp->field = value;
  ```

4. **Minimal risk**: The fix doesn't change any logic or functionality.
   It only changes how the structure is initialized, making it extremely
   unlikely to introduce regressions.

5. **Precedent from similar commits**: Looking at the historical commits
   marked "YES" for backporting:
   - Similar Commit #1: Reduced stack frame in qed driver using
     `noinline_for_stack`
   - Similar Commit #4: Reduced stack usage in ethtool with clang using
     `noinline_for_stack`

   Both addressed the same class of problem (excessive stack usage with
clang) and were considered suitable for stable.

6. **Measurable improvement**: The stack usage reduction from 1472 to
   208 bytes is dramatic and well-documented by the kernel test robot,
   providing clear evidence of the fix's effectiveness.

The commit meets the stable kernel criteria of fixing an important bug
with minimal risk and a contained change. While it doesn't explicitly
include a "Cc: stable" tag, the nature of the fix (preventing potential
stack overflow) makes it a good candidate for stable backporting.

 arch/um/drivers/vector_kern.c | 42 +++++++++++------------------------
 1 file changed, 13 insertions(+), 29 deletions(-)

diff --git a/arch/um/drivers/vector_kern.c b/arch/um/drivers/vector_kern.c
index b97bb52dd5626..70f8d7e87fb81 100644
--- a/arch/um/drivers/vector_kern.c
+++ b/arch/um/drivers/vector_kern.c
@@ -1592,35 +1592,19 @@ static void vector_eth_configure(
 
 	device->dev = dev;
 
-	*vp = ((struct vector_private)
-		{
-		.list			= LIST_HEAD_INIT(vp->list),
-		.dev			= dev,
-		.unit			= n,
-		.options		= get_transport_options(def),
-		.rx_irq			= 0,
-		.tx_irq			= 0,
-		.parsed			= def,
-		.max_packet		= get_mtu(def) + ETH_HEADER_OTHER,
-		/* TODO - we need to calculate headroom so that ip header
-		 * is 16 byte aligned all the time
-		 */
-		.headroom		= get_headroom(def),
-		.form_header		= NULL,
-		.verify_header		= NULL,
-		.header_rxbuffer	= NULL,
-		.header_txbuffer	= NULL,
-		.header_size		= 0,
-		.rx_header_size		= 0,
-		.rexmit_scheduled	= false,
-		.opened			= false,
-		.transport_data		= NULL,
-		.in_write_poll		= false,
-		.coalesce		= 2,
-		.req_size		= get_req_size(def),
-		.in_error		= false,
-		.bpf			= NULL
-	});
+	INIT_LIST_HEAD(&vp->list);
+	vp->dev		= dev;
+	vp->unit	= n;
+	vp->options	= get_transport_options(def);
+	vp->parsed	= def;
+	vp->max_packet	= get_mtu(def) + ETH_HEADER_OTHER;
+	/*
+	 * TODO - we need to calculate headroom so that ip header
+	 * is 16 byte aligned all the time
+	 */
+	vp->headroom	= get_headroom(def);
+	vp->coalesce	= 2;
+	vp->req_size	= get_req_size(def);
 
 	dev->features = dev->hw_features = (NETIF_F_SG | NETIF_F_FRAGLIST);
 	INIT_WORK(&vp->reset_tx, vector_reset_tx);
-- 
2.39.5

[PATCH AUTOSEL 6.15 20/23] riscv: vdso: Exclude .rodata from the PT_DYNAMIC segment

[Sasha Levin]

From: Fangrui Song <i@maskray.me>

[ Upstream commit e0eb1b6b0cd29ca7793c501d5960fd36ba11f110 ]

.rodata is implicitly included in the PT_DYNAMIC segment due to
inheriting the segment of the preceding .dynamic section (in both GNU ld
and LLD).  When the .rodata section's size is not a multiple of 16
bytes on riscv64, llvm-readelf will report a "PT_DYNAMIC dynamic table
is invalid" warning.  Note: in the presence of the .dynamic section, GNU
readelf and llvm-readelf's -d option decodes the dynamic section using
the section.

This issue arose after commit 8f8c1ff879fab60f80f3a7aec3000f47e5b03ba9
("riscv: vdso.lds.S: remove hardcoded 0x800 .text start addr"), which
placed .rodata directly after .dynamic by removing .eh_frame.

This patch resolves the implicit inclusion into PT_DYNAMIC by explicitly
specifying the :text output section phdr.

Reported-by: Nathan Chancellor <nathan@kernel.org>
Closes: https://github.com/ClangBuiltLinux/linux/issues/2093
Signed-off-by: Fangrui Song <i@maskray.me>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Link: https://lore.kernel.org/r/20250602-riscv-vdso-v1-1-0620cf63cff0@maskray.me
Signed-off-by: Palmer Dabbelt <palmer@dabbelt.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---

**YES**

This commit should be backported to stable kernel trees for the
following reasons:

1. **Regression Fix**: The commit explicitly fixes a regression
   introduced by commit 8f8c1ff879fab60f80f3a7aec3000f47e5b03ba9
   ("riscv: vdso.lds.S: remove hardcoded 0x800 .text start addr").
   Regression fixes are high-priority candidates for stable backporting.

2. **Fixes Invalid ELF Structure**: The issue is not merely cosmetic.
   The PT_DYNAMIC segment is reported as "invalid" by llvm-readelf,
   indicating the ELF structure is technically incorrect. From the
   linker script:
  ```
  .dynamic        : { *(.dynamic) }               :text   :dynamic
  .rodata         : {
  ...
  }  // implicitly inherits :dynamic from previous section
  ```
  Without the explicit `:text` specification, `.rodata` inherits the
  segment attributes from the preceding `.dynamic` section, incorrectly
  including it in the PT_DYNAMIC segment.

3. **Minimal and Safe Change**: The fix is a single-line change that
   adds `:text` to the `.rodata` section definition:
  ```diff
   - }
   +    }                                               :text
   ```
   This explicitly places `.rodata` in the text segment only, excluding
it from PT_DYNAMIC. The change has no functional impact beyond
correcting the ELF structure.

4. **Similar to Previous Backported Commits**: This is similar to commit
   #2 in the reference list ("riscv: vdso: fix section overlapping under
   some conditions") which was marked YES for backporting. Both commits
   fix structural issues in the vDSO linker script that cause
   build/tooling errors.

5. **Potential for Broader Impact**: While the immediate symptom is a
   warning from llvm-readelf, an invalid PT_DYNAMIC segment could
   potentially cause issues with:
   - Build systems that validate ELF structures
   - Runtime loaders with strict ELF validation
   - Debugging and analysis tools
   - Future toolchain versions that may be less tolerant of invalid
     structures

The commit meets the stable tree criteria: it fixes an important bug
(invalid ELF structure), the fix is small and contained, and there's
minimal risk of regression.

 arch/riscv/kernel/vdso/vdso.lds.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/riscv/kernel/vdso/vdso.lds.S b/arch/riscv/kernel/vdso/vdso.lds.S
index 8e86965a8aae4..646e268ede443 100644
--- a/arch/riscv/kernel/vdso/vdso.lds.S
+++ b/arch/riscv/kernel/vdso/vdso.lds.S
@@ -30,7 +30,7 @@ SECTIONS
 		*(.data .data.* .gnu.linkonce.d.*)
 		*(.dynbss)
 		*(.bss .bss.* .gnu.linkonce.b.*)
-	}
+	}						:text
 
 	.note		: { *(.note.*) }		:text	:note
 
-- 
2.39.5