Le lundi 26 octobre 2009 04:24:42, James Harper a écrit :
> > I think that run-windows-in-a-vm is a better idea than using BartPE,
> > now if Bacula can restore acl stream from BackupWrite/Read, i would be
> > very happy too!
>
> The patch tacked onto the end of this email accomplishes this, with the
> following (serious) restrictions/bugs:
>
> . Doesn't seem to work on directories. I assume that this is because
> when directories get restored, processWin32BackupAPIBlock is never
> called at any point... does that sound right? Any suggestions as to how
> to work around it?
This call is present on many places (bextrat, filed/restore.c, etc...), perhaps
it is missing at some places, but it's hard to say.
> . Some other type of file is failing to apply ACL's... can I get the
> filename from the jcr? Or any other way?
Marco is responsible for all ACL stuffs, he is doing a great job to cleanup all
this part, i think that once you will have the good approach, he will have
some comments on how integrate your changes.
> . I'm restoring the Extended Attributes too, but haven't yet tested if
> they actually do anything
> . Needs the current 'Advanced' version of ntfs-3g
If it's done from a LiveCD, this is just a documentation problem.
> . The patch as it stands assumes that you are restoring to a filesystem
> mounted with the 'Advanced' version of ntfs-3g. If you were to try and
> restore to any other filesystem you would fill your logs with errors
> about failing to set attributes.
This is where Marco will help us, the current architecture doesn't have this
problem.
> . Alternate Data Streams are not restored. Easy enough to add I think,
> but ntfs-3g has a few different ways of allowing access to them and I
> don't know which is best at this point.
What can we expect from this Stream ?
> . If a win32 backup stream (other than the normal data) crossed a bacula
> buffer boundary then it won't work. The alternative is to read the
> stream into a memory buffer and only set the xattr when it is completely
> read into memory. I haven't seen that happen yet but I'm pretty sure
> that it could.
We can wait 2 days to get Kern advises on this particular subject, i think
that we call the BackupRead without any buffer boundary, so it shouldn't cause
problem.
> . Reparse points aren't restored, but I think that's just another 3
> lines of code.
Great!
> . Needs the attr libraries
For sure, this is already needed for all other extended attributes.
> . Assumes that you are restoring onto the same endianness as you backed
> up on. Windows is and has always been little-endian only so I don't
> think this is a problem. To make it endian independent would be a
> nightmare.
We can assume it at this time, in any cases, it needs careful testing.
Great idea, great job, thanks you!
Bye
> James
>
> diff --git a/bacula/src/findlib/bfile.c b/bacula/src/findlib/bfile.c
> index 3a98480..4950e93 100644
> --- a/bacula/src/findlib/bfile.c
> +++ b/bacula/src/findlib/bfile.c
> @@ -38,6 +38,7 @@
>
> #include "bacula.h"
> #include "find.h"
> +#include <attr/xattr.h>
>
> const int dbglvl = 200;
>
> @@ -244,6 +245,10 @@ bool processWin32BackupAPIBlock (BFILE *bfd, void
> *pBuffer, ssize_t dwSize)
> */
> int32_t dwSizeHeader = 20;
>
> + if (pContext->bIsInData && pContext->header_stream.dwStreamId !=
> WIN32_BACKUP_DATA) {
> + Jmsg(bfd->jcr, M_WARNING, 0, _("WIN32_BACKUP_%d split across
> buffers\n"), pContext->header_stream.dwStreamId);
> + }
> +
> do {
> if (pContext->liNextHeader >= dwSize) {
> dwDataLen = dwSize-dwDataOffset;
> @@ -256,8 +261,21 @@ bool processWin32BackupAPIBlock (BFILE *bfd, void
> *pBuffer, ssize_t dwSize)
> /* flush */
> /* copy block of real DATA */
> if (pContext->bIsInData) {
> - if (bwrite(bfd, ((char *)pBuffer)+dwDataOffset, dwDataLen) !=
> (ssize_t)dwDataLen)
> - return false;
> + switch (pContext->header_stream.dwStreamId)
> + {
> + case WIN32_BACKUP_DATA:
> + if (bwrite(bfd, ((char *)pBuffer)+dwDataOffset, dwDataLen)
> != (ssize_t)dwDataLen)
> + return false;
> + break;
> + case WIN32_BACKUP_EA_DATA:
> + if (fsetxattr(bfd->fid, "system.ntfs_attrib", ((char
> *)pBuffer)+dwDataOffset, dwDataLen, 0) != 0)
> + Jmsg(bfd->jcr, M_WARNING, 0, _("failed to set EA on fd
> %d length %d for '%s' - %s\n"), bfd->fid, dwDataLen, "somefile",
> strerror(errno));
> + break;
> + case WIN32_BACKUP_SECURITY_DATA:
> + if (fsetxattr(bfd->fid, "system.ntfs_acl", ((char
> *)pBuffer)+dwDataOffset, dwDataLen, 0) != 0)
> + Jmsg(bfd->jcr, M_WARNING, 0, _("failed to set ACL on fd
> %d length %d for '%s' - %s\n"), bfd->fid, dwDataLen, "somefile",
> strerror(errno));
> + break;
> + }
> }
>
> if (pContext->liNextHeader < dwSize) {/* is a header in this
> block ? */
> @@ -304,7 +322,7 @@ bool processWin32BackupAPIBlock (BFILE *bfd, void
> *pBuffer, ssize_t dwSize)
> int64_LE2BE (&(pContext->liNextHeader),
> pContext->header_stream.Size);
> pContext->liNextHeader += dwDataOffset;
>
> - pContext->bIsInData = pContext->header_stream.dwStreamId ==
> WIN32_BACKUP_DATA;
> + pContext->bIsInData = true;
> if (dwDataOffset == dwSize)
> bContinue = false;
> } else {
> diff --git a/bacula/src/findlib/bfile.h b/bacula/src/findlib/bfile.h
> index b22d50c..9575147 100644
> --- a/bacula/src/findlib/bfile.h
> +++ b/bacula/src/findlib/bfile.h
> @@ -43,6 +43,8 @@
> */
>
> #define WIN32_BACKUP_DATA 1
> +#define WIN32_BACKUP_EA_DATA 2
> +#define WIN32_BACKUP_SECURITY_DATA 3
>
> typedef struct _BWIN32_STREAM_ID {
> int32_t dwStreamId;
>
------------------------------------------------------------------------------
Come build with us! The BlackBerry(R) Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay
ahead of the curve. Join us from November 9 - 12, 2009. Register now!
http://p.sf.net/sfu/devconference
_______________________________________________
Bacula-devel mailing list
Bacula-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-devel
