* [PATCH 0/2] ExtPkg Updates
@ 2021-09-09 20:40 Jeff Brasen
2021-09-09 20:41 ` [PATCH 1/2] Ext4Pkg: Improve Binding support behavior Jeff Brasen
` (2 more replies)
0 siblings, 3 replies; 8+ messages in thread
From: Jeff Brasen @ 2021-09-09 20:40 UTC (permalink / raw)
To: devel; +Cc: pedro.falcato, Jeff Brasen
I have been using the new Ext4Pkg and been pretty successful and it is solving a use case we had.
Had a couple updates to propose
1. Changed the implementation of the binding protocol to both check if the driver is already bound
to the partition as well as added a really quick check to validate the magic value in supported.
This improves performance when you have a large number of non-ext4 partitions on the system.
2. As we are planning on using this for boot support we want to support unclean filesystem states in
case the user doesn't reset cleanly. I added a check if the recovery journal is present and if so treat
the filesystem as read-only (I know the driver is only RO at this point, but figured if you added write
support prior to recovery journal support we would want that). With this everything seems to work great.
I can add this under a FeaturePcd if desired as well.
Change log
v1 - Initial revision
Jeff Brasen (2):
Ext4Pkg: Improve Binding support behavior
Ext4Pkg: Support non-cleanlty unmounted filesystems
Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h | 14 +++++++
Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c | 57 +++++++++++++++++++++------
Features/Ext4Pkg/Ext4Dxe/Superblock.c | 45 +++++++++++++++++++--
3 files changed, 99 insertions(+), 17 deletions(-)
--
2.17.1
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 1/2] Ext4Pkg: Improve Binding support behavior
2021-09-09 20:40 [PATCH 0/2] ExtPkg Updates Jeff Brasen
@ 2021-09-09 20:41 ` Jeff Brasen
2021-09-10 4:22 ` Pedro Falcato
2021-09-10 16:24 ` [edk2-devel] " Marvin Häuser
2021-09-09 20:41 ` [PATCH 2/2] Ext4Pkg: Support non-cleanlty unmounted filesystems Jeff Brasen
2021-09-10 4:40 ` [PATCH 0/2] ExtPkg Updates Pedro Falcato
2 siblings, 2 replies; 8+ messages in thread
From: Jeff Brasen @ 2021-09-09 20:41 UTC (permalink / raw)
To: devel; +Cc: pedro.falcato, Jeff Brasen
A couple improvements to improve performance.
Add check to return ACCESS_DENIED if already connected
Add check to verify superblock magic during supported to deduce start calls
Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
---
Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h | 14 +++++++
Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c | 57 +++++++++++++++++++++------
Features/Ext4Pkg/Ext4Dxe/Superblock.c | 34 ++++++++++++++++
3 files changed, 92 insertions(+), 13 deletions(-)
diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
index 64eab455db..9a3938e671 100644
--- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
+++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
@@ -1117,4 +1117,18 @@ Ext4GetVolumeName (
OUT UINTN *VolNameLen
);
+/**
+ Checks only superblock magic value.
+
+ @param[in] DiskIo Pointer to the DiskIo.
+ @param[in] BlockIo Pointer to the BlockIo.
+
+ @return TRUE if a valid ext4 superblock, else FALSE.
+**/
+BOOLEAN
+Ext4SuperblockCheckMagic (
+ IN EFI_DISK_IO_PROTOCOL *DiskIo,
+ IN EFI_BLOCK_IO_PROTOCOL *BlockIo
+ );
+
#endif
diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
index ea2e048d77..cb1e6d532a 100644
--- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
+++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
@@ -631,7 +631,6 @@ Ext4Unload (
@retval EFI_ACCESS_DENIED The device specified by ControllerHandle and
RemainingDevicePath is already being managed by a different
driver or an application that requires exclusive access.
- Currently not implemented.
@retval EFI_UNSUPPORTED The device specified by ControllerHandle and
RemainingDevicePath is not supported by the driver specified by This.
**/
@@ -643,32 +642,64 @@ Ext4IsBindingSupported (
IN EFI_DEVICE_PATH *RemainingDevicePath OPTIONAL
)
{
- // Note to self: EFI_OPEN_PROTOCOL_TEST_PROTOCOL lets us not close the
- // protocol and ignore the output argument entirely
-
- EFI_STATUS Status;
+ EFI_STATUS Status;
+ EFI_DISK_IO_PROTOCOL *DiskIo = NULL;
+ EFI_BLOCK_IO_PROTOCOL *BlockIo = NULL;
+ //
+ // Open the IO Abstraction(s) needed to perform the supported test
+ //
Status = gBS->OpenProtocol (
ControllerHandle,
&gEfiDiskIoProtocolGuid,
- NULL,
- BindingProtocol->ImageHandle,
+ (VOID **) &DiskIo,
+ BindingProtocol->DriverBindingHandle,
ControllerHandle,
- EFI_OPEN_PROTOCOL_TEST_PROTOCOL
+ EFI_OPEN_PROTOCOL_BY_DRIVER
);
if (EFI_ERROR (Status)) {
- return Status;
+ goto Exit;
}
-
+ //
+ // Open the IO Abstraction(s) needed to perform the supported test
+ //
Status = gBS->OpenProtocol (
ControllerHandle,
&gEfiBlockIoProtocolGuid,
- NULL,
- BindingProtocol->ImageHandle,
+ (VOID **) &BlockIo,
+ BindingProtocol->DriverBindingHandle,
ControllerHandle,
- EFI_OPEN_PROTOCOL_TEST_PROTOCOL
+ EFI_OPEN_PROTOCOL_GET_PROTOCOL
);
+ if (EFI_ERROR (Status)) {
+ goto Exit;
+ }
+
+ if (!Ext4SuperblockCheckMagic (DiskIo, BlockIo)) {
+ Status = EFI_UNSUPPORTED;
+ }
+
+Exit:
+ //
+ // Close the I/O Abstraction(s) used to perform the supported test
+ //
+ if (DiskIo != NULL) {
+ gBS->CloseProtocol (
+ ControllerHandle,
+ &gEfiDiskIoProtocolGuid,
+ BindingProtocol->DriverBindingHandle,
+ ControllerHandle
+ );
+ }
+ if (BlockIo != NULL) {
+ gBS->CloseProtocol (
+ ControllerHandle,
+ &gEfiBlockIoProtocolGuid,
+ BindingProtocol->DriverBindingHandle,
+ ControllerHandle
+ );
+ }
return Status;
}
diff --git a/Features/Ext4Pkg/Ext4Dxe/Superblock.c b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
index c321d8c3d8..1f8cdd3705 100644
--- a/Features/Ext4Pkg/Ext4Dxe/Superblock.c
+++ b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
@@ -34,6 +34,40 @@ STATIC CONST UINT32 gSupportedIncompatFeat =
// this is desired, it's fairly trivial to look for EFI_VOLUME_CORRUPTED
// references and add some Ext4SignalCorruption function + function call.
+/**
+ Checks only superblock magic value.
+
+ @param[in] DiskIo Pointer to the DiskIo.
+ @param[in] BlockIo Pointer to the BlockIo.
+
+ @return TRUE if a valid ext4 superblock, else FALSE.
+**/
+BOOLEAN
+Ext4SuperblockCheckMagic (
+ IN EFI_DISK_IO_PROTOCOL *DiskIo,
+ IN EFI_BLOCK_IO_PROTOCOL *BlockIo
+ )
+{
+ UINT16 Magic;
+ EFI_STATUS Status;
+
+ Status = DiskIo->ReadDisk (DiskIo,
+ BlockIo->Media->MediaId,
+ EXT4_SUPERBLOCK_OFFSET + OFFSET_OF (EXT4_SUPERBLOCK, s_magic),
+ sizeof (Magic),
+ &Magic
+ );
+ if (EFI_ERROR (Status)) {
+ return FALSE;
+ }
+
+ if (Magic != EXT4_SIGNATURE) {
+ return FALSE;
+ }
+
+ return TRUE;
+}
+
/**
Does brief validation of the ext4 superblock.
--
2.17.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH 2/2] Ext4Pkg: Support non-cleanlty unmounted filesystems
2021-09-09 20:40 [PATCH 0/2] ExtPkg Updates Jeff Brasen
2021-09-09 20:41 ` [PATCH 1/2] Ext4Pkg: Improve Binding support behavior Jeff Brasen
@ 2021-09-09 20:41 ` Jeff Brasen
2021-09-10 4:35 ` Pedro Falcato
2021-09-10 4:40 ` [PATCH 0/2] ExtPkg Updates Pedro Falcato
2 siblings, 1 reply; 8+ messages in thread
From: Jeff Brasen @ 2021-09-09 20:41 UTC (permalink / raw)
To: devel; +Cc: pedro.falcato, Jeff Brasen
Support for uncleanly mounted filesystems, if there is a recovery
journal mark filesystem as read-only.
Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
---
Features/Ext4Pkg/Ext4Dxe/Superblock.c | 11 +++++++----
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/Features/Ext4Pkg/Ext4Dxe/Superblock.c b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
index 1f8cdd3705..444dd3ca97 100644
--- a/Features/Ext4Pkg/Ext4Dxe/Superblock.c
+++ b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
@@ -18,7 +18,7 @@ STATIC CONST UINT32 gSupportedIncompatFeat =
EXT4_FEATURE_INCOMPAT_64BIT | EXT4_FEATURE_INCOMPAT_DIRDATA |
EXT4_FEATURE_INCOMPAT_FLEX_BG | EXT4_FEATURE_INCOMPAT_FILETYPE |
EXT4_FEATURE_INCOMPAT_EXTENTS | EXT4_FEATURE_INCOMPAT_LARGEDIR |
- EXT4_FEATURE_INCOMPAT_MMP;
+ EXT4_FEATURE_INCOMPAT_MMP | EXT4_FEATURE_INCOMPAT_RECOVER;
// Future features that may be nice additions in the future:
// 1) Btree support: Required for write support and would speed up lookups in large directories.
@@ -88,10 +88,8 @@ Ext4SuperblockValidate (
return FALSE;
}
- // Is this correct behaviour? Imagine the power cuts out, should the system fail to boot because
- // we're scared of touching something corrupt?
if ((Sb->s_state & EXT4_FS_STATE_UNMOUNTED) == 0) {
- return FALSE;
+ DEBUG ((DEBUG_WARN, "[ext4] filesystem was not unmounted cleanly\n"));
}
return TRUE;
@@ -214,6 +212,11 @@ Ext4OpenSuperblock (
return EFI_UNSUPPORTED;
}
+ if ((Partition->FeaturesIncompat & EXT4_FEATURE_INCOMPAT_RECOVER) == EXT4_FEATURE_INCOMPAT_RECOVER) {
+ DEBUG ((DEBUG_WARN, "[ext4] Needs journal recovery mount read-only\n"));
+ Partition->ReadOnly = TRUE;
+ }
+
// At the time of writing, it's the only supported checksum.
if (Partition->FeaturesCompat & EXT4_FEATURE_RO_COMPAT_METADATA_CSUM &&
Sb->s_checksum_type != EXT4_CHECKSUM_CRC32C) {
--
2.17.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH 1/2] Ext4Pkg: Improve Binding support behavior
2021-09-09 20:41 ` [PATCH 1/2] Ext4Pkg: Improve Binding support behavior Jeff Brasen
@ 2021-09-10 4:22 ` Pedro Falcato
2021-09-10 4:34 ` Jeff Brasen
2021-09-10 16:24 ` [edk2-devel] " Marvin Häuser
1 sibling, 1 reply; 8+ messages in thread
From: Pedro Falcato @ 2021-09-10 4:22 UTC (permalink / raw)
To: Jeff Brasen; +Cc: edk2-devel-groups-io
Hi Jeff,
Comments below.
The patch itself looks good.
Nitpick on the commit message: I'd reword the commit message to
something like "Improve Ext4IsBindingSupported() behavior"
or "Improve IsBindingSupported behavior", since this patch doesn't
change Ext4Bind(), which does the actual bind.
On Thu, Sep 9, 2021 at 9:41 PM Jeff Brasen <jbrasen@nvidia.com> wrote:
>
> A couple improvements to improve performance.
> Add check to return ACCESS_DENIED if already connected
> Add check to verify superblock magic during supported to deduce start calls
>
> Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
> ---
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h | 14 +++++++
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c | 57 +++++++++++++++++++++------
> Features/Ext4Pkg/Ext4Dxe/Superblock.c | 34 ++++++++++++++++
> 3 files changed, 92 insertions(+), 13 deletions(-)
>
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> index 64eab455db..9a3938e671 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> +++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> @@ -1117,4 +1117,18 @@ Ext4GetVolumeName (
> OUT UINTN *VolNameLen
> );
>
> +/**
> + Checks only superblock magic value.
> +
> + @param[in] DiskIo Pointer to the DiskIo.
> + @param[in] BlockIo Pointer to the BlockIo.
> +
> + @return TRUE if a valid ext4 superblock, else FALSE.
> +**/
> +BOOLEAN
> +Ext4SuperblockCheckMagic (
> + IN EFI_DISK_IO_PROTOCOL *DiskIo,
> + IN EFI_BLOCK_IO_PROTOCOL *BlockIo
> + );
I'd reword the first part of the comment into something like "Checks
the superblock's magic value.".
Aligning the two "Pointer to the ...Io" would be neat :)
Types and parameter names need to be separated by at least two
columns. So, something like:
IN EFI_DISK_IO_PROTOCOL *DiskIo,
IN EFI_BLOCK_IO_PROTOCOL *BlockIo
The feedback above also applies to the function's definition below, in
Superblock.c.
> +
> #endif
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> index ea2e048d77..cb1e6d532a 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> +++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> @@ -631,7 +631,6 @@ Ext4Unload (
> @retval EFI_ACCESS_DENIED The device specified by ControllerHandle and
> RemainingDevicePath is already being managed by a different
> driver or an application that requires exclusive access.
> - Currently not implemented.
> @retval EFI_UNSUPPORTED The device specified by ControllerHandle and
> RemainingDevicePath is not supported by the driver specified by This.
> **/
> @@ -643,32 +642,64 @@ Ext4IsBindingSupported (
> IN EFI_DEVICE_PATH *RemainingDevicePath OPTIONAL
> )
> {
> - // Note to self: EFI_OPEN_PROTOCOL_TEST_PROTOCOL lets us not close the
> - // protocol and ignore the output argument entirely
> -
> - EFI_STATUS Status;
> + EFI_STATUS Status;
> + EFI_DISK_IO_PROTOCOL *DiskIo = NULL;
> + EFI_BLOCK_IO_PROTOCOL *BlockIo = NULL;
>
Initialization of variables are always separate from the declaration. Example:
EFI_DISK_IO_PROTOCOL *DiskIo;
...
DiskIo = NULL;
> + //
> + // Open the IO Abstraction(s) needed to perform the supported test
> + //
> Status = gBS->OpenProtocol (
> ControllerHandle,
> &gEfiDiskIoProtocolGuid,
> - NULL,
> - BindingProtocol->ImageHandle,
> + (VOID **) &DiskIo,
> + BindingProtocol->DriverBindingHandle,
> ControllerHandle,
> - EFI_OPEN_PROTOCOL_TEST_PROTOCOL
> + EFI_OPEN_PROTOCOL_BY_DRIVER
> );
>
> if (EFI_ERROR (Status)) {
> - return Status;
> + goto Exit;
> }
> -
> + //
> + // Open the IO Abstraction(s) needed to perform the supported test
> + //
> Status = gBS->OpenProtocol (
> ControllerHandle,
> &gEfiBlockIoProtocolGuid,
> - NULL,
> - BindingProtocol->ImageHandle,
> + (VOID **) &BlockIo,
> + BindingProtocol->DriverBindingHandle,
> ControllerHandle,
> - EFI_OPEN_PROTOCOL_TEST_PROTOCOL
> + EFI_OPEN_PROTOCOL_GET_PROTOCOL
> );
> + if (EFI_ERROR (Status)) {
> + goto Exit;
> + }
> +
> + if (!Ext4SuperblockCheckMagic (DiskIo, BlockIo)) {
> + Status = EFI_UNSUPPORTED;
> + }
> +
> +Exit:
> + //
> + // Close the I/O Abstraction(s) used to perform the supported test
> + //
> + if (DiskIo != NULL) {
> + gBS->CloseProtocol (
> + ControllerHandle,
> + &gEfiDiskIoProtocolGuid,
> + BindingProtocol->DriverBindingHandle,
> + ControllerHandle
> + );
> + }
> + if (BlockIo != NULL) {
> + gBS->CloseProtocol (
> + ControllerHandle,
> + &gEfiBlockIoProtocolGuid,
> + BindingProtocol->DriverBindingHandle,
> + ControllerHandle
> + );
> + }
> return Status;
> }
>
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Superblock.c b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> index c321d8c3d8..1f8cdd3705 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> +++ b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> @@ -34,6 +34,40 @@ STATIC CONST UINT32 gSupportedIncompatFeat =
> // this is desired, it's fairly trivial to look for EFI_VOLUME_CORRUPTED
> // references and add some Ext4SignalCorruption function + function call.
>
> +/**
> + Checks only superblock magic value.
> +
> + @param[in] DiskIo Pointer to the DiskIo.
> + @param[in] BlockIo Pointer to the BlockIo.
> +
> + @return TRUE if a valid ext4 superblock, else FALSE.
> +**/
> +BOOLEAN
> +Ext4SuperblockCheckMagic (
> + IN EFI_DISK_IO_PROTOCOL *DiskIo,
> + IN EFI_BLOCK_IO_PROTOCOL *BlockIo
> + )
Same as above.
> +{
> + UINT16 Magic;
> + EFI_STATUS Status;
> +
> + Status = DiskIo->ReadDisk (DiskIo,
> + BlockIo->Media->MediaId,
> + EXT4_SUPERBLOCK_OFFSET + OFFSET_OF (EXT4_SUPERBLOCK, s_magic),
> + sizeof (Magic),
> + &Magic
> + );
Splitting a function call in multiple lines should be done like in
https://edk2-docs.gitbook.io/edk-ii-c-coding-standards-specification/5_source_files/52_spacing#5-2-2-4-subsequent-lines-of-multi-line-function-calls-should-line-up-two-spaces-from-the-beginning-of-the-function-name.
> + if (EFI_ERROR (Status)) {
> + return FALSE;
> + }
> +
> + if (Magic != EXT4_SIGNATURE) {
> + return FALSE;
> + }
> +
> + return TRUE;
> +}
> +
> /**
> Does brief validation of the ext4 superblock.
>
> --
> 2.17.1
>
--
Pedro Falcato
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 1/2] Ext4Pkg: Improve Binding support behavior
2021-09-10 4:22 ` Pedro Falcato
@ 2021-09-10 4:34 ` Jeff Brasen
0 siblings, 0 replies; 8+ messages in thread
From: Jeff Brasen @ 2021-09-10 4:34 UTC (permalink / raw)
To: Pedro Falcato; +Cc: edk2-devel-groups-io
[-- Attachment #1: Type: text/plain, Size: 7952 bytes --]
Sounds good will update that in a v2 patch tomorrow
Get Outlook for Android<https://aka.ms/ghei36>
________________________________
From: Pedro Falcato <pedro.falcato@gmail.com>
Sent: Thursday, September 9, 2021 10:22:51 PM
To: Jeff Brasen <jbrasen@nvidia.com>
Cc: edk2-devel-groups-io <devel@edk2.groups.io>
Subject: Re: [PATCH 1/2] Ext4Pkg: Improve Binding support behavior
External email: Use caution opening links or attachments
Hi Jeff,
Comments below.
The patch itself looks good.
Nitpick on the commit message: I'd reword the commit message to
something like "Improve Ext4IsBindingSupported() behavior"
or "Improve IsBindingSupported behavior", since this patch doesn't
change Ext4Bind(), which does the actual bind.
On Thu, Sep 9, 2021 at 9:41 PM Jeff Brasen <jbrasen@nvidia.com> wrote:
>
> A couple improvements to improve performance.
> Add check to return ACCESS_DENIED if already connected
> Add check to verify superblock magic during supported to deduce start calls
>
> Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
> ---
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h | 14 +++++++
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c | 57 +++++++++++++++++++++------
> Features/Ext4Pkg/Ext4Dxe/Superblock.c | 34 ++++++++++++++++
> 3 files changed, 92 insertions(+), 13 deletions(-)
>
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> index 64eab455db..9a3938e671 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> +++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> @@ -1117,4 +1117,18 @@ Ext4GetVolumeName (
> OUT UINTN *VolNameLen
> );
>
> +/**
> + Checks only superblock magic value.
> +
> + @param[in] DiskIo Pointer to the DiskIo.
> + @param[in] BlockIo Pointer to the BlockIo.
> +
> + @return TRUE if a valid ext4 superblock, else FALSE.
> +**/
> +BOOLEAN
> +Ext4SuperblockCheckMagic (
> + IN EFI_DISK_IO_PROTOCOL *DiskIo,
> + IN EFI_BLOCK_IO_PROTOCOL *BlockIo
> + );
I'd reword the first part of the comment into something like "Checks
the superblock's magic value.".
Aligning the two "Pointer to the ...Io" would be neat :)
Types and parameter names need to be separated by at least two
columns. So, something like:
IN EFI_DISK_IO_PROTOCOL *DiskIo,
IN EFI_BLOCK_IO_PROTOCOL *BlockIo
The feedback above also applies to the function's definition below, in
Superblock.c.
> +
> #endif
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> index ea2e048d77..cb1e6d532a 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> +++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> @@ -631,7 +631,6 @@ Ext4Unload (
> @retval EFI_ACCESS_DENIED The device specified by ControllerHandle and
> RemainingDevicePath is already being managed by a different
> driver or an application that requires exclusive access.
> - Currently not implemented.
> @retval EFI_UNSUPPORTED The device specified by ControllerHandle and
> RemainingDevicePath is not supported by the driver specified by This.
> **/
> @@ -643,32 +642,64 @@ Ext4IsBindingSupported (
> IN EFI_DEVICE_PATH *RemainingDevicePath OPTIONAL
> )
> {
> - // Note to self: EFI_OPEN_PROTOCOL_TEST_PROTOCOL lets us not close the
> - // protocol and ignore the output argument entirely
> -
> - EFI_STATUS Status;
> + EFI_STATUS Status;
> + EFI_DISK_IO_PROTOCOL *DiskIo = NULL;
> + EFI_BLOCK_IO_PROTOCOL *BlockIo = NULL;
>
Initialization of variables are always separate from the declaration. Example:
EFI_DISK_IO_PROTOCOL *DiskIo;
...
DiskIo = NULL;
> + //
> + // Open the IO Abstraction(s) needed to perform the supported test
> + //
> Status = gBS->OpenProtocol (
> ControllerHandle,
> &gEfiDiskIoProtocolGuid,
> - NULL,
> - BindingProtocol->ImageHandle,
> + (VOID **) &DiskIo,
> + BindingProtocol->DriverBindingHandle,
> ControllerHandle,
> - EFI_OPEN_PROTOCOL_TEST_PROTOCOL
> + EFI_OPEN_PROTOCOL_BY_DRIVER
> );
>
> if (EFI_ERROR (Status)) {
> - return Status;
> + goto Exit;
> }
> -
> + //
> + // Open the IO Abstraction(s) needed to perform the supported test
> + //
> Status = gBS->OpenProtocol (
> ControllerHandle,
> &gEfiBlockIoProtocolGuid,
> - NULL,
> - BindingProtocol->ImageHandle,
> + (VOID **) &BlockIo,
> + BindingProtocol->DriverBindingHandle,
> ControllerHandle,
> - EFI_OPEN_PROTOCOL_TEST_PROTOCOL
> + EFI_OPEN_PROTOCOL_GET_PROTOCOL
> );
> + if (EFI_ERROR (Status)) {
> + goto Exit;
> + }
> +
> + if (!Ext4SuperblockCheckMagic (DiskIo, BlockIo)) {
> + Status = EFI_UNSUPPORTED;
> + }
> +
> +Exit:
> + //
> + // Close the I/O Abstraction(s) used to perform the supported test
> + //
> + if (DiskIo != NULL) {
> + gBS->CloseProtocol (
> + ControllerHandle,
> + &gEfiDiskIoProtocolGuid,
> + BindingProtocol->DriverBindingHandle,
> + ControllerHandle
> + );
> + }
> + if (BlockIo != NULL) {
> + gBS->CloseProtocol (
> + ControllerHandle,
> + &gEfiBlockIoProtocolGuid,
> + BindingProtocol->DriverBindingHandle,
> + ControllerHandle
> + );
> + }
> return Status;
> }
>
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Superblock.c b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> index c321d8c3d8..1f8cdd3705 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> +++ b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> @@ -34,6 +34,40 @@ STATIC CONST UINT32 gSupportedIncompatFeat =
> // this is desired, it's fairly trivial to look for EFI_VOLUME_CORRUPTED
> // references and add some Ext4SignalCorruption function + function call.
>
> +/**
> + Checks only superblock magic value.
> +
> + @param[in] DiskIo Pointer to the DiskIo.
> + @param[in] BlockIo Pointer to the BlockIo.
> +
> + @return TRUE if a valid ext4 superblock, else FALSE.
> +**/
> +BOOLEAN
> +Ext4SuperblockCheckMagic (
> + IN EFI_DISK_IO_PROTOCOL *DiskIo,
> + IN EFI_BLOCK_IO_PROTOCOL *BlockIo
> + )
Same as above.
> +{
> + UINT16 Magic;
> + EFI_STATUS Status;
> +
> + Status = DiskIo->ReadDisk (DiskIo,
> + BlockIo->Media->MediaId,
> + EXT4_SUPERBLOCK_OFFSET + OFFSET_OF (EXT4_SUPERBLOCK, s_magic),
> + sizeof (Magic),
> + &Magic
> + );
Splitting a function call in multiple lines should be done like in
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fedk2-docs.gitbook.io%2Fedk-ii-c-coding-standards-specification%2F5_source_files%2F52_spacing%235-2-2-4-subsequent-lines-of-multi-line-function-calls-should-line-up-two-spaces-from-the-beginning-of-the-function-name&data=04%7C01%7Cjbrasen%40nvidia.com%7C6e1a40882279499fbb0008d97412af02%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0%7C637668446481896854%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=WZjbNs2Q3eU6QH%2F1RZbxakJnhVYOE4%2Bt%2BWF68JragVg%3D&reserved=0.
> + if (EFI_ERROR (Status)) {
> + return FALSE;
> + }
> +
> + if (Magic != EXT4_SIGNATURE) {
> + return FALSE;
> + }
> +
> + return TRUE;
> +}
> +
> /**
> Does brief validation of the ext4 superblock.
>
> --
> 2.17.1
>
--
Pedro Falcato
[-- Attachment #2: Type: text/html, Size: 15286 bytes --]
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 2/2] Ext4Pkg: Support non-cleanlty unmounted filesystems
2021-09-09 20:41 ` [PATCH 2/2] Ext4Pkg: Support non-cleanlty unmounted filesystems Jeff Brasen
@ 2021-09-10 4:35 ` Pedro Falcato
0 siblings, 0 replies; 8+ messages in thread
From: Pedro Falcato @ 2021-09-10 4:35 UTC (permalink / raw)
To: Jeff Brasen; +Cc: edk2-devel-groups-io
Comments below.
The patch itself also looks good.
Commit message issue: typo on "non-cleanlty".
On Thu, Sep 9, 2021 at 9:41 PM Jeff Brasen <jbrasen@nvidia.com> wrote:
>
> Support for uncleanly mounted filesystems, if there is a recovery
> journal mark filesystem as read-only.
>
> Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
> ---
> Features/Ext4Pkg/Ext4Dxe/Superblock.c | 11 +++++++----
> 1 file changed, 7 insertions(+), 4 deletions(-)
>
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Superblock.c b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> index 1f8cdd3705..444dd3ca97 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> +++ b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> @@ -18,7 +18,7 @@ STATIC CONST UINT32 gSupportedIncompatFeat =
> EXT4_FEATURE_INCOMPAT_64BIT | EXT4_FEATURE_INCOMPAT_DIRDATA |
> EXT4_FEATURE_INCOMPAT_FLEX_BG | EXT4_FEATURE_INCOMPAT_FILETYPE |
> EXT4_FEATURE_INCOMPAT_EXTENTS | EXT4_FEATURE_INCOMPAT_LARGEDIR |
> - EXT4_FEATURE_INCOMPAT_MMP;
> + EXT4_FEATURE_INCOMPAT_MMP | EXT4_FEATURE_INCOMPAT_RECOVER;
>
> // Future features that may be nice additions in the future:
> // 1) Btree support: Required for write support and would speed up lookups in large directories.
> @@ -88,10 +88,8 @@ Ext4SuperblockValidate (
> return FALSE;
> }
>
> - // Is this correct behaviour? Imagine the power cuts out, should the system fail to boot because
> - // we're scared of touching something corrupt?
> if ((Sb->s_state & EXT4_FS_STATE_UNMOUNTED) == 0) {
> - return FALSE;
> + DEBUG ((DEBUG_WARN, "[ext4] filesystem was not unmounted cleanly\n"));
Nitpick: filesystem should be capitalized.
> }
>
> return TRUE;
> @@ -214,6 +212,11 @@ Ext4OpenSuperblock (
> return EFI_UNSUPPORTED;
> }
>
> + if ((Partition->FeaturesIncompat & EXT4_FEATURE_INCOMPAT_RECOVER) == EXT4_FEATURE_INCOMPAT_RECOVER) {
New code that wants to test for features/feature-sets should use
EXT4_HAS_INCOMPAT, EXT4_HAS_COMPAT, EXT4_HAS_RO_COMPAT.
The code in this function that's testing for features using i.e:
FeaturesIncompat & FEATURE manually should likely be replaced by the
macros.
> + DEBUG ((DEBUG_WARN, "[ext4] Needs journal recovery mount read-only\n"));
The debug message looks poorly worded; something like "[ext4] Needs
journal recovery, mounting read-only\n" sounds good?
> + Partition->ReadOnly = TRUE;
> + }
> +
> // At the time of writing, it's the only supported checksum.
> if (Partition->FeaturesCompat & EXT4_FEATURE_RO_COMPAT_METADATA_CSUM &&
> Sb->s_checksum_type != EXT4_CHECKSUM_CRC32C) {
> --
> 2.17.1
>
--
Pedro Falcato
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 0/2] ExtPkg Updates
2021-09-09 20:40 [PATCH 0/2] ExtPkg Updates Jeff Brasen
2021-09-09 20:41 ` [PATCH 1/2] Ext4Pkg: Improve Binding support behavior Jeff Brasen
2021-09-09 20:41 ` [PATCH 2/2] Ext4Pkg: Support non-cleanlty unmounted filesystems Jeff Brasen
@ 2021-09-10 4:40 ` Pedro Falcato
2 siblings, 0 replies; 8+ messages in thread
From: Pedro Falcato @ 2021-09-10 4:40 UTC (permalink / raw)
To: Jeff Brasen; +Cc: edk2-devel-groups-io
Hi Jeff,
Thanks for the patches! It's great that Ext4Pkg is already getting used!
I've looked at each patch and replied with feedback.
In general, they look good, although there are some minor issues. I
hadn't thought of the 1st patch but the 2nd seemed inevitable
for real world systems that may not unmount cleanly.
Best regards,
Pedro
On Thu, Sep 9, 2021 at 9:41 PM Jeff Brasen <jbrasen@nvidia.com> wrote:
>
> I have been using the new Ext4Pkg and been pretty successful and it is solving a use case we had.
>
> Had a couple updates to propose
>
> 1. Changed the implementation of the binding protocol to both check if the driver is already bound
> to the partition as well as added a really quick check to validate the magic value in supported.
> This improves performance when you have a large number of non-ext4 partitions on the system.
>
> 2. As we are planning on using this for boot support we want to support unclean filesystem states in
> case the user doesn't reset cleanly. I added a check if the recovery journal is present and if so treat
> the filesystem as read-only (I know the driver is only RO at this point, but figured if you added write
> support prior to recovery journal support we would want that). With this everything seems to work great.
> I can add this under a FeaturePcd if desired as well.
>
> Change log
>
> v1 - Initial revision
>
> Jeff Brasen (2):
> Ext4Pkg: Improve Binding support behavior
> Ext4Pkg: Support non-cleanlty unmounted filesystems
>
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h | 14 +++++++
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c | 57 +++++++++++++++++++++------
> Features/Ext4Pkg/Ext4Dxe/Superblock.c | 45 +++++++++++++++++++--
> 3 files changed, 99 insertions(+), 17 deletions(-)
>
> --
> 2.17.1
>
--
Pedro Falcato
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [edk2-devel] [PATCH 1/2] Ext4Pkg: Improve Binding support behavior
2021-09-09 20:41 ` [PATCH 1/2] Ext4Pkg: Improve Binding support behavior Jeff Brasen
2021-09-10 4:22 ` Pedro Falcato
@ 2021-09-10 16:24 ` Marvin Häuser
1 sibling, 0 replies; 8+ messages in thread
From: Marvin Häuser @ 2021-09-10 16:24 UTC (permalink / raw)
To: devel, jbrasen; +Cc: pedro.falcato
Good day,
On 09/09/2021 22:41, Jeff Brasen via groups.io wrote:
> A couple improvements to improve performance.
> Add check to return ACCESS_DENIED if already connected
This "performance improvement" actually aligns the load behaviour with
the UEFI spec, maybe it should be mentioned?
> Add check to verify superblock magic during supported to deduce start calls
>
> Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
> ---
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h | 14 +++++++
> Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c | 57 +++++++++++++++++++++------
> Features/Ext4Pkg/Ext4Dxe/Superblock.c | 34 ++++++++++++++++
> 3 files changed, 92 insertions(+), 13 deletions(-)
>
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> index 64eab455db..9a3938e671 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> +++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h
> @@ -1117,4 +1117,18 @@ Ext4GetVolumeName (
> OUT UINTN *VolNameLen
> );
>
> +/**
> + Checks only superblock magic value.
> +
> + @param[in] DiskIo Pointer to the DiskIo.
> + @param[in] BlockIo Pointer to the BlockIo.
> +
> + @return TRUE if a valid ext4 superblock, else FALSE.
> +**/
> +BOOLEAN
> +Ext4SuperblockCheckMagic (
> + IN EFI_DISK_IO_PROTOCOL *DiskIo,
> + IN EFI_BLOCK_IO_PROTOCOL *BlockIo
> + );
> +
> #endif
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> index ea2e048d77..cb1e6d532a 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> +++ b/Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c
> @@ -631,7 +631,6 @@ Ext4Unload (
> @retval EFI_ACCESS_DENIED The device specified by ControllerHandle and
> RemainingDevicePath is already being managed by a different
> driver or an application that requires exclusive access.
> - Currently not implemented.
> @retval EFI_UNSUPPORTED The device specified by ControllerHandle and
> RemainingDevicePath is not supported by the driver specified by This.
> **/
> @@ -643,32 +642,64 @@ Ext4IsBindingSupported (
> IN EFI_DEVICE_PATH *RemainingDevicePath OPTIONAL
> )
> {
> - // Note to self: EFI_OPEN_PROTOCOL_TEST_PROTOCOL lets us not close the
> - // protocol and ignore the output argument entirely
> -
> - EFI_STATUS Status;
> + EFI_STATUS Status;
> + EFI_DISK_IO_PROTOCOL *DiskIo = NULL;
> + EFI_BLOCK_IO_PROTOCOL *BlockIo = NULL;
>
> + //
> + // Open the IO Abstraction(s) needed to perform the supported test
> + //
> Status = gBS->OpenProtocol (
> ControllerHandle,
> &gEfiDiskIoProtocolGuid,
> - NULL,
> - BindingProtocol->ImageHandle,
> + (VOID **) &DiskIo,
> + BindingProtocol->DriverBindingHandle,
> ControllerHandle,
> - EFI_OPEN_PROTOCOL_TEST_PROTOCOL
> + EFI_OPEN_PROTOCOL_BY_DRIVER
> );
>
> if (EFI_ERROR (Status)) {
> - return Status;
> + goto Exit;
This change is not required as nothing was opened on failure; see below.
> }
> -
> + //
> + // Open the IO Abstraction(s) needed to perform the supported test
> + //
> Status = gBS->OpenProtocol (
> ControllerHandle,
> &gEfiBlockIoProtocolGuid,
> - NULL,
> - BindingProtocol->ImageHandle,
> + (VOID **) &BlockIo,
> + BindingProtocol->DriverBindingHandle,
> ControllerHandle,
> - EFI_OPEN_PROTOCOL_TEST_PROTOCOL
> + EFI_OPEN_PROTOCOL_GET_PROTOCOL
Why do the open modes mismatch if both protocols are equally stored in
the private partition data? Is the owner of Block I/O a different one
than the one of Disk I/O?
> );
> + if (EFI_ERROR (Status)) {
> + goto Exit;
> + }
> +
> + if (!Ext4SuperblockCheckMagic (DiskIo, BlockIo)) {
> + Status = EFI_UNSUPPORTED;
> + }
This can easily be rewritten to not require "goto":
if (!EFI_ERROR (Status) && !Ext4SuperblockCheckMagic (DiskIo, BlockIo)) {
Status = EFI_UNSUPPORTED;
}
With the change above this allows to drop the label and any "goto" code.
> +
> +Exit:
> + //
> + // Close the I/O Abstraction(s) used to perform the supported test
> + //
> + if (DiskIo != NULL) {
> + gBS->CloseProtocol (
> + ControllerHandle,
> + &gEfiDiskIoProtocolGuid,
> + BindingProtocol->DriverBindingHandle,
> + ControllerHandle
> + );
> + }
> + if (BlockIo != NULL) {
> + gBS->CloseProtocol (
> + ControllerHandle,
> + &gEfiBlockIoProtocolGuid,
> + BindingProtocol->DriverBindingHandle,
> + ControllerHandle
> + );
> + }
Protocols retrieved by GET_PROTOCOL are not closed.
> return Status;
> }
>
> diff --git a/Features/Ext4Pkg/Ext4Dxe/Superblock.c b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> index c321d8c3d8..1f8cdd3705 100644
> --- a/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> +++ b/Features/Ext4Pkg/Ext4Dxe/Superblock.c
> @@ -34,6 +34,40 @@ STATIC CONST UINT32 gSupportedIncompatFeat =
> // this is desired, it's fairly trivial to look for EFI_VOLUME_CORRUPTED
> // references and add some Ext4SignalCorruption function + function call.
>
> +/**
> + Checks only superblock magic value.
Whether it "only" checks the magic is more of a functional detail than
an actual description.
> +
> + @param[in] DiskIo Pointer to the DiskIo.
> + @param[in] BlockIo Pointer to the BlockIo.
> +
> + @return TRUE if a valid ext4 superblock, else FALSE.
Maybe use "@returns Whether the partition has a valid EXT4 superblock
magic" for readability?
Best regards,
Marvin
> +**/
> +BOOLEAN
> +Ext4SuperblockCheckMagic (
> + IN EFI_DISK_IO_PROTOCOL *DiskIo,
> + IN EFI_BLOCK_IO_PROTOCOL *BlockIo
> + )
> +{
> + UINT16 Magic;
> + EFI_STATUS Status;
> +
> + Status = DiskIo->ReadDisk (DiskIo,
> + BlockIo->Media->MediaId,
> + EXT4_SUPERBLOCK_OFFSET + OFFSET_OF (EXT4_SUPERBLOCK, s_magic),
> + sizeof (Magic),
> + &Magic
> + );
> + if (EFI_ERROR (Status)) {
> + return FALSE;
> + }
> +
> + if (Magic != EXT4_SIGNATURE) {
> + return FALSE;
> + }
> +
> + return TRUE;
> +}
> +
> /**
> Does brief validation of the ext4 superblock.
>
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2021-09-10 16:25 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2021-09-09 20:40 [PATCH 0/2] ExtPkg Updates Jeff Brasen
2021-09-09 20:41 ` [PATCH 1/2] Ext4Pkg: Improve Binding support behavior Jeff Brasen
2021-09-10 4:22 ` Pedro Falcato
2021-09-10 4:34 ` Jeff Brasen
2021-09-10 16:24 ` [edk2-devel] " Marvin Häuser
2021-09-09 20:41 ` [PATCH 2/2] Ext4Pkg: Support non-cleanlty unmounted filesystems Jeff Brasen
2021-09-10 4:35 ` Pedro Falcato
2021-09-10 4:40 ` [PATCH 0/2] ExtPkg Updates Pedro Falcato
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox