JAGNYA DATTA TRIPATHY has uploaded this change for review.
NFSv4: fix OPEN_DOWNGRADE validation for BOTH after separate opens
When a client opens a file twice -- first with READ-only, then with
WRITE-only -- the server accumulates share_access_prev as:
(1 << OPEN4_SHARE_ACCESS_READ) = bit 1 (value 2)
(1 << OPEN4_SHARE_ACCESS_WRITE) = bit 2 (value 4)
share_access_prev = 6
The combined share_access is now BOTH (3). A subsequent
OPEN_DOWNGRADE to BOTH is a valid no-op: the client holds all rights
that BOTH represents. However the existing check:
share_access_prev & (1 << OPEN4_SHARE_ACCESS_BOTH)
= 6 & (1 << 3) = 6 & 8 = 0 -> NFS4ERR_INVAL
fails because BOTH (3) was never passed as a single mode value to any
OPEN call, so its bit (bit 3 = value 8) was never stored.
RFC 7530 §16.19.4 / RFC 5661 §18.18.3 state that the downgrade target
must be "a subset of those used in all OPEN calls". For BOTH this
means: valid if BOTH was directly opened OR if both READ and WRITE
were individually opened (their union equals BOTH).
Introduce share_downgrade_allowed() to implement this rule. For any
target other than BOTH the function reduces to the original single-bit
check, preserving all existing pynfs OPDG test behaviour:
OPDG2: OPEN(BOTH) -> downgrade(READ) -- target=READ != BOTH
original check: 8 & 2 = 0 -> NFS4ERR_INVAL (correct, unchanged)
OPDG3: OPEN(WRITE)+OPEN(BOTH) -> downgrade(READ) -- target=READ
original check: 12 & 2 = 0 -> NFS4ERR_INVAL (correct, unchanged)
The BOTH special case only activates for the accumulation scenario and
correctly returns NFS4_OK where the original code returned NFS4ERR_INVAL.
Change-Id: I58751d89b50933db5223e3620f23d1857893ca77
Signed-off-by: Jagnya Tripathy <jagnya.tripathy@gmail.com>
---
M src/Protocols/NFS/nfs4_op_open_downgrade.c
1 file changed, 30 insertions(+), 5 deletions(-)
git pull ssh://review.gerrithub.io:29418/ffilz/nfs-ganesha refs/changes/22/1238122/1
To view, visit change 1238122. To unsubscribe, or for help writing mail filters, visit settings.