[PATCH V2] kho: Convert error handling to immediate return pattern

[PATCH V2] kho: Convert error handling to immediate return pattern

[LongWei27]

From: Long Wei <longwei27@huawei.com>

Replace the chained error accumulation (err |= func()) with immediate
per-step error checking and early return in kho_out_fdt_setup().

Two problems with the current approach:
1. It continues executing FDT operations even after a failure, which
   is pointless and potentially unsafe on an invalid FDT context.
2. Bitwise OR of multiple negative error codes produces a result that
   may not represent any actual error, making debugging misleading.

Early return fails fast and preserves the original error code for
accurate diagnosis.

No functional change intended.

Signed-off-by: Long Wei <longwei27@huawei.com>
---
 kernel/liveupdate/kexec_handover.c | 32 +++++++++++++++++++++---------
 1 file changed, 23 insertions(+), 9 deletions(-)

diff --git a/kernel/liveupdate/kexec_handover.c b/kernel/liveupdate/kexec_handover.c
index 175c08a6e..93443ad84 100644
--- a/kernel/liveupdate/kexec_handover.c
+++ b/kernel/liveupdate/kexec_handover.c
@@ -1438,20 +1438,34 @@ static __init int kho_out_fdt_setup(void)
 	int err;
 
 	err = fdt_create(root, PAGE_SIZE);
-	err |= fdt_finish_reservemap(root);
-	err |= fdt_begin_node(root, "");
-	err |= fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
+	if (err)
+		return err;
+	err = fdt_finish_reservemap(root);
+	if (err)
+		return err;
+	err = fdt_begin_node(root, "");
+	if (err)
+		return err;
+	err = fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
+	if (err)
+		return err;
 
 	preserved_mem_tree_pa = virt_to_phys(tree->root);
 
-	err |= fdt_property(root, KHO_FDT_MEMORY_MAP_PROP_NAME,
-			    &preserved_mem_tree_pa,
-			    sizeof(preserved_mem_tree_pa));
+	err = fdt_property(root, KHO_FDT_MEMORY_MAP_PROP_NAME,
+			   &preserved_mem_tree_pa,
+			   sizeof(preserved_mem_tree_pa));
+	if (err)
+		return err;
 
-	err |= fdt_end_node(root);
-	err |= fdt_finish(root);
+	err = fdt_end_node(root);
+	if (err)
+		return err;
+	err = fdt_finish(root);
+	if (err)
+		return err;
 
-	return err;
+	return 0;
 }
 
 static void __init kho_in_kexec_metadata(void)
-- 
2.43.0

Re: [PATCH V2] kho: Convert error handling to immediate return pattern

[Matthew Wilcox]

On Mon, Aug 24, 2026 at 10:44:48AM +0800, LongWei27 wrote:
> +++ b/kernel/liveupdate/kexec_handover.c
> @@ -1438,20 +1438,34 @@ static __init int kho_out_fdt_setup(void)
>  	int err;
>  
>  	err = fdt_create(root, PAGE_SIZE);
> -	err |= fdt_finish_reservemap(root);
> -	err |= fdt_begin_node(root, "");
> -	err |= fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
> +	if (err)
> +		return err;
> +	err = fdt_finish_reservemap(root);
> +	if (err)
> +		return err;
> +	err = fdt_begin_node(root, "");
> +	if (err)
> +		return err;
> +	err = fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
> +	if (err)
> +		return err;

Less verbose:

 	err = fdt_create(root, PAGE_SIZE);
-	err |= fdt_finish_reservemap(root);
-	err |= fdt_begin_node(root, "");
-	err |= fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
+	if (!err)
+		err = fdt_finish_reservemap(root);
+	if (!err)
+		err = fdt_begin_node(root, "");
+	if (!err)
+		err = fdt_property_string(root, "compatible",
+				KHO_FDT_COMPATIBLE);

up to the maintainer which one they prefer ...

Re: [PATCH V2] kho: Convert error handling to immediate return pattern

[longwei (I)]

Hi Matthew,

Thanks for the review.

I chose the explicit `if (err) return err` pattern for clarity and
"fail fast" behavior. However, I'm happy to adopt the chained pattern
if the maintainer prefers it.

Thanks,
Long Wei

在 2026/8/24 10:54, Matthew Wilcox 写道:
> On Mon, Aug 24, 2026 at 10:44:48AM +0800, LongWei27 wrote:
>> +++ b/kernel/liveupdate/kexec_handover.c
>> @@ -1438,20 +1438,34 @@ static __init int kho_out_fdt_setup(void)
>>  	int err;
>>  
>>  	err = fdt_create(root, PAGE_SIZE);
>> -	err |= fdt_finish_reservemap(root);
>> -	err |= fdt_begin_node(root, "");
>> -	err |= fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
>> +	if (err)
>> +		return err;
>> +	err = fdt_finish_reservemap(root);
>> +	if (err)
>> +		return err;
>> +	err = fdt_begin_node(root, "");
>> +	if (err)
>> +		return err;
>> +	err = fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
>> +	if (err)
>> +		return err;
> 
> Less verbose:
> 
>  	err = fdt_create(root, PAGE_SIZE);
> -	err |= fdt_finish_reservemap(root);
> -	err |= fdt_begin_node(root, "");
> -	err |= fdt_property_string(root, "compatible", KHO_FDT_COMPATIBLE);
> +	if (!err)
> +		err = fdt_finish_reservemap(root);
> +	if (!err)
> +		err = fdt_begin_node(root, "");
> +	if (!err)
> +		err = fdt_property_string(root, "compatible",
> +				KHO_FDT_COMPATIBLE);
> 
> up to the maintainer which one they prefer ...

Re: [PATCH V2] kho: Convert error handling to immediate return pattern

[Mike Rapoport]

Hi,


> Replace the chained error accumulation (err |= func()) with immediate
> per-step error checking and early return in kho_out_fdt_setup().
> 
> Two problems with the current approach:
> 1. It continues executing FDT operations even after a failure, which
>    is pointless and potentially unsafe on an invalid FDT context.

Did you actually see failures here or was it an LLM suggesting that they
may happen?

> 2. Bitwise OR of multiple negative error codes produces a result that
>    may not represent any actual error, making debugging misleading.
> 
> Early return fails fast and preserves the original error code for
> accurate diagnosis.

The caller does not care about the exact error value, so the diagnosis
statement is moot at best.
Moreover, libfdt does not return Linux error values.

-- 
Sincerely yours,
Mike.

Re: [PATCH V2] kho: Convert error handling to immediate return pattern

[longwei (I)]

Hi Mike Rapoport，

Thanks for the review.
To be honest, I haven't observed any actual failures in this path in our production environment.
This change was identified during our internal team code review—the err |= func() pattern looked concerning to us at first glance, so we attempted to clean it up.

However, I agree with your assessment: the caller doesn't care about the exact error value, libfdt uses its own error codes, and this is a cold path where failures are unlikely. The change is indeed unnecessary churn without a concrete bug to fix.

I'll drop this patch. Please ignore this submission.

Thanks,
Long Wei

在 2026/8/31 19:28, Mike Rapoport 写道:
> Hi,
> 
> 
>> Replace the chained error accumulation (err |= func()) with immediate
>> per-step error checking and early return in kho_out_fdt_setup().
>>
>> Two problems with the current approach:
>> 1. It continues executing FDT operations even after a failure, which
>>    is pointless and potentially unsafe on an invalid FDT context.
> 
> Did you actually see failures here or was it an LLM suggesting that they
> may happen?
> 
>> 2. Bitwise OR of multiple negative error codes produces a result that
>>    may not represent any actual error, making debugging misleading.
>>
>> Early return fails fast and preserves the original error code for
>> accurate diagnosis.
> 
> The caller does not care about the exact error value, so the diagnosis
> statement is moot at best.
> Moreover, libfdt does not return Linux error values.
>