[PATCH] efivarfs: Never return ENOENT from firmware again

[PATCH] efivarfs: Never return ENOENT from firmware again

[Lingzhu Xiang]

Previously in 1fa7e69 efi_status_to_err() translated firmware status
EFI_NOT_FOUND to -EIO instead of -ENOENT for efivarfs operations to
avoid confusion. After refactoring in e14ab23, it is also used in other
places where the translation may be unnecessary.

So move the translation to efivarfs specific code. Also return EOF
for reading zero-length files, which is what users would expect.

Cc: Josh Boyer <jwboyer@redhat.com>
Cc: Jeremy Kerr <jk@ozlabs.org>
Cc: Lee, Chun-Yi <jlee@suse.com>
Cc: Andy Whitcroft <apw@canonical.com>
Signed-off-by: Lingzhu Xiang <lxiang@redhat.com>
---
This is to be applied against mainline or Matt Fleming's chainsaw branch.

 fs/efivarfs/file.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/fs/efivarfs/file.c b/fs/efivarfs/file.c
index bfb5315..8dd524f 100644
--- a/fs/efivarfs/file.c
+++ b/fs/efivarfs/file.c
@@ -44,8 +44,11 @@ static ssize_t efivarfs_file_write(struct file *file,
 
 	bytes = efivar_entry_set_get_size(var, attributes, &datasize,
 					  data, &set);
-	if (!set && bytes)
+	if (!set && bytes) {
+		if (bytes == -ENOENT)
+			bytes = -EIO;
 		goto out;
+	}
 
 	if (bytes == -ENOENT) {
 		drop_nlink(inode);
@@ -76,7 +79,14 @@ static ssize_t efivarfs_file_read(struct file *file, char __user *userbuf,
 	int err;
 
 	err = efivar_entry_size(var, &datasize);
-	if (err)
+
+	/*
+	 * efivarfs represents uncommitted variables with
+	 * zero-length files. Reading them should return EOF.
+	 */
+	if (err == -ENOENT)
+		return 0;
+	else if (err)
 		return err;
 
 	data = kmalloc(datasize + sizeof(attributes), GFP_KERNEL);
-- 
1.7.11.7

Re: [PATCH] efivarfs: Never return ENOENT from firmware again

[Matt Fleming]

On 05/10/2013 11:29 AM, Lingzhu Xiang wrote:
> Previously in 1fa7e69 efi_status_to_err() translated firmware status
> EFI_NOT_FOUND to -EIO instead of -ENOENT for efivarfs operations to
> avoid confusion. After refactoring in e14ab23, it is also used in other
> places where the translation may be unnecessary.
> 
> So move the translation to efivarfs specific code. Also return EOF
> for reading zero-length files, which is what users would expect.
> 
> Cc: Josh Boyer <jwboyer@redhat.com>
> Cc: Jeremy Kerr <jk@ozlabs.org>
> Cc: Lee, Chun-Yi <jlee@suse.com>
> Cc: Andy Whitcroft <apw@canonical.com>
> Signed-off-by: Lingzhu Xiang <lxiang@redhat.com>
> ---
> This is to be applied against mainline or Matt Fleming's chainsaw branch.

Applied, thanks.

FYI, I rebased my 'urgent' branch on v3.10-rc1, so your patch now
applies there too.

Re: [PATCH] efivarfs: Never return ENOENT from firmware again

[joeyli]

Hi Matt, 

於 一，2013-05-13 於 20:21 +0100，Matt Fleming 提到：
> On 05/10/2013 11:29 AM, Lingzhu Xiang wrote:
> > Previously in 1fa7e69 efi_status_to_err() translated firmware status
> > EFI_NOT_FOUND to -EIO instead of -ENOENT for efivarfs operations to
> > avoid confusion. After refactoring in e14ab23, it is also used in other
> > places where the translation may be unnecessary.
> > 
> > So move the translation to efivarfs specific code. Also return EOF
> > for reading zero-length files, which is what users would expect.
> > 
> > Cc: Josh Boyer <jwboyer@redhat.com>
> > Cc: Jeremy Kerr <jk@ozlabs.org>
> > Cc: Lee, Chun-Yi <jlee@suse.com>
> > Cc: Andy Whitcroft <apw@canonical.com>
> > Signed-off-by: Lingzhu Xiang <lxiang@redhat.com>
> > ---
> > This is to be applied against mainline or Matt Fleming's chainsaw branch.
> 
> Applied, thanks.
> 
> FYI, I rebased my 'urgent' branch on v3.10-rc1, so your patch now
> applies there too.
> 
> 

I found the "[PATCH] x86, efi: initial the local variable of DataSize to
zero" lost in urgent branch, and it also didn't in Linus's v3.10-rc1.

Did this patch move to other branch?


Thanks a lot!
Joey Lee

Re: [PATCH] efivarfs: Never return ENOENT from firmware again

[Matt Fleming]

On 05/14/2013 05:36 AM, joeyli wrote:
> I found the "[PATCH] x86, efi: initial the local variable of DataSize to
> zero" lost in urgent branch, and it also didn't in Linus's v3.10-rc1.
> 
> Did this patch move to other branch?

It appears I dropped this patch when doing the rebase against -rc1. I'm
really sorry about that. I've applied it again.

Thanks for being so diligent.

Re: [PATCH] efivarfs: Never return ENOENT from firmware again

[joeyli]

於 二，2013-05-14 於 08:16 +0100，Matt Fleming 提到：
> On 05/14/2013 05:36 AM, joeyli wrote:
> > I found the "[PATCH] x86, efi: initial the local variable of DataSize to
> > zero" lost in urgent branch, and it also didn't in Linus's v3.10-rc1.
> > 
> > Did this patch move to other branch?
> 
> It appears I dropped this patch when doing the rebase against -rc1. I'm
> really sorry about that. I've applied it again.

Thanks for your help!

> 
> Thanks for being so diligent.
> --

That because I need apply this patch on v3.9 and v3.10 when debugging
issue on Acer machine.


Thank a lot!
Joey Lee