| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Remove mutex_lock from probe in order to avoid the following warning:
[ 8.526613] Freeing unused kernel memory: 232K (c0683000 - c06bd000)
starting pid 56, tty '': '/etc/rc.d/rcS'
[ 9.110314]
[ 9.111864] =====================================
[ 9.116603] [ BUG: init/1 still has locks held! ]
[ 9.121488] 3.15.0-rc4-next-20140509-00001-g319564e #1154 Not tainted
[ 9.128071] -------------------------------------
[ 9.132825] 1 lock held by init/1:
[ 9.136252] #0: (global_mutex){+.+.+.}, at: [<c0387d68>] mxs_dcp_probe+0x14
[ 9.144196]
[ 9.144196] stack backtrace:
[ 9.148888] CPU: 0 PID: 1 Comm: init Not tainted 3.15.0-rc4-next-20140509-004
[ 9.157610] [<c000da40>] (unwind_backtrace) from [<c000bda4>] (show_stack+0x)
[ 9.165595] [<c000bda4>] (show_stack) from [<c00153d4>] (do_fork+0x2c8/0x3cc)
[ 9.172921] [<c00153d4>] (do_fork) from [<c0015550>] (sys_vfork+0x20/0x2c)
[ 9.179973] [<c0015550>] (sys_vfork) from [<c0009580>] (ret_fast_syscall+0x0)
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Re-initialize keys_fit_inline to avoid using its stale encrypt() shared
descriptor value prior to building descriptors for the decrypt() and
givencrypt() cases.
Signed-off-by: Vakul Garg <vakul@freescale.com>
[reworded commit text, enhanced code readability]
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
As we are preparing to enable multiplatform support on EXYNOS,
we can no longer include mach/*.h or plat/*.h headers from device
drivers.
The s5p-sss driver was just enabled for EXYNOS when it used to
be used only on s5pv210, and it includes two samsung platform
specific header files for historic reasons. Fortunately, it no
longer actually needs them, so we can remove the #includes and
avoid the problem
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Cc: Tomasz Figa <t.figa@samsung.com>
Cc: <linux-samsung-soc@vger.kernel.org>
Cc: Vladimir Zapolskiy <vz@mleia.com>
Reviewed-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Currently, the driver enqueues a request only if the busy bit is
false. And every request initiates a dequeue. If 2 requests arrive
simultaneously, only one of them will be dequeued.
To avoid this senario, we will enqueue the next request irrespective
of the system condition (that is what queue is here for). Also
schedule at a tasklet immediatly after the current request is done.
The tasklet will dequeue the next request in the queue, giving
continuous loop. tasklet will exit if there are no requests in the
queue.
Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
CC: David S. Miller <davem@davemloft.net>
CC: <linux-samsung-soc@vger.kernel.org>
Acked-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch set adds use of clk_prepare/clk_unprepare as
required by generic clock framework.
Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
CC: David S. Miller <davem@davemloft.net>
CC: <linux-samsung-soc@vger.kernel.org>
Acked-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch adds code to validate "iv" buffer before trying to
memcpy the contents
Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
CC: David S. Miller <davem@davemloft.net>
CC: <linux-samsung-soc@vger.kernel.org>
Acked-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch modifies Kconfig such that ARCH_EXYNOS SoCs
which includes (Exynos4210, Exynos5250 and Exynos5420)
can also select Samsung SSS(Security SubSystem) driver.
Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
CC: David S. Miller <davem@davemloft.net>
CC: <linux-samsung-soc@vger.kernel.org>
Acked-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch adds new compatible and variant struct to support the SSS
module on Exynos4 (Exynos4210), Exynos5 (Exynos5420 and Exynos5250)
for which
1. AES register are at an offset of 0x200 and
2. hash interrupt is not available
Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
CC: David S. Miller <davem@davemloft.net>
CC: <linux-samsung-soc@vger.kernel.org>
Acked-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch adds device tree support to the s5p-sss.c crypto driver.
Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
CC: David S. Miller <davem@davemloft.net>
CC: <linux-samsung-soc@vger.kernel.org>
Acked-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch uses the platform_get_irq() instead of the
platform_get_irq_byname(). Making feeder control interrupt
as resource "0" and hash interrupt as "1".
reasons for this change.
1. Cannot find any Arch which is currently using this driver
2. Samsung Exynos4 and 5 SoCs only use the feeder control interrupt
3. Patches adding support for DT and H/W version are in pipeline
Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
CC: David S. Miller <davem@davemloft.net>
CC: <linux-samsung-soc@vger.kernel.org>
Acked-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Fix the checkpatch warnings that the strings were split across
multiple lines. Checkpatch now complains about lines over 80,
but this is better, since we can actually grep the source code
for these strings now.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Clean up the remnants from the rework. Constify function arguments.
Note that checkpatch again complains about this space before newline,
but this is the original code behavior, so I'm keeping it.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This macro is just like an encyclopedia of string handling done wrong.
This must die. This is so wrong on so many levels.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Clean this function up and rework it into sensible shape. This function
now contains one single dev_err() instead of the previous insanity full
of memory allocation, chaotic string handling and use of SPRINTFCAT().
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Clean this function up and rework it into sensible shape. This function
now contains one single dev_err() instead of the previous insanity full
of memory allocation, possible stack overwriting, chaotic string handling
and use of SPRINTFCAT().
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Just dissolve this function so it's not in the way of applying
further white magic cleanup down the line.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Fix the functions which can be obviously done right with a simple
dev_err() now. While at it, further press the on-stack allocation
of buffer for sprintf() voodoo down into the abominated functions.
This patch cleans up most of the functions and leaves just two
remaining functions, report_ccb_status() and report_deco_status()
ugly and unhappy.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Pass the error type string into the functions, so they can handle
the printing of the string. This is now still using the very unsafe
sprintf(), but we will fix that.
While at this, pass the device pointer too, so we can dev_err()
functions readily when we start fixing this proper.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Implement fast-path error code printout for errors with no associated
handler function. This reduces calls to this kmalloc() nonsense in
SPRINTFCAT() already.
Note that the format of output is compatible with the old code, even
if -- exposed like this -- it looks a bit weird. Checkpatch complains
on this one as well.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Pull the error code <-> error string mapping tables out of the function
so the code becomes readable. This lets me see the real flesh of the
functions, without all that flab clouding the view.
Note: There is a checkpatch issue with quoted strings across multiple
lines. I will fix that in a subsequent patch to keep the changes
small and separate.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
The tentacles of this function were firmly attached to various
places in the CAAM code. Just cut them, or this cthulhu function
will sprout them anew.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
The length shoule be 64 bit alignment and the block size shoule be 64 bit in aes cfb64 mode.
Signed-off-by: Leilei Zhao <leilei.zhao@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
The block size of aes cfb8 mode shoule be 8 bit.
Signed-off-by: Leilei Zhao <leilei.zhao@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
In case hash key is bigger than algorithm block size, it is hashed.
In this case, memory is allocated to keep this hash in hashed_key.
hashed_key has to be freed on the key_dma dma mapping error path.
Signed-off-by: Horia Geanta <horia.geanta@freescale.com>
Reviewed-by: Marek Vasut <marex@denx.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
GFP_ATOMIC memory allocation could fail.
In this case, avoid NULL pointer dereference and notify user.
Cc: <stable@vger.kernel.org> # 3.2+
Cc: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
pm_runtime_get_sync may not always succeed depending on SoC involved.
So handle the error appropriately ensuring usage_count is accurate in
case of failure.
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-by: Felipe Balbi <balbi@ti.com>
Reported-by: Joachim Eastwood <manabian@gmail.com>
Tested-by: Joachim Eastwood <manabian@gmail.com>
Acked-by: Joel Fernandes <joelf@ti.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
As result of deprecation of MSI-X/MSI enablement functions
pci_enable_msix() and pci_enable_msi_block() all drivers
using these two interfaces need to be updated to use the
new pci_enable_msi_range() or pci_enable_msi_exact()
and pci_enable_msix_range() or pci_enable_msix_exact()
interfaces.
Signed-off-by: Alexander Gordeev <agordeev@redhat.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: linux-crypto@vger.kernel.org
Cc: linux-pci@vger.kernel.org
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
dma_map_single
- The 4-byte sg_mid_buf is located in the middle of the coherence memory
sg_cpu. Don't call dma_map_single to get its physical address. Get the its
base physical address from the physical address of sg_cpu instead.
- Should set up the dma descriptor data after the 4-byte sg_mid_buf is
filled in completely from next sg buffer.
- memory copy from sg buffer should be done via virtual address.
- Remove unused reference to blackfin header
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Reviewed-by: Marek Vasut <marex@denx.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
in case of multiple crc devices are probed.
Call platform_set_drvdata() before adding new CRC device into the list.
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Reviewed-by: Marek Vasut <marex@denx.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Move architecture independant crc header file out of the blackfin folder.
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Reviewed-by: Marek Vasut <marex@denx.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Use devm_ioremap_resource() because devm_request_and_ioremap() is
obsoleted by devm_ioremap_resource().
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Joel Fernandes <joelf@ti.com>
Reviewed-by: Marek Vasut <marex@denx.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch replaces rcu_assign_pointer(x, NULL) with RCU_INIT_POINTER(x, NULL)
The rcu_assign_pointer() ensures that the initialization of a structure
is carried out before storing a pointer to that structure.
And in the case of the NULL pointer, there is no structure to initialize.
So, rcu_assign_pointer(p, NULL) can be safely converted to RCU_INIT_POINTER(p, NULL)
Signed-off-by: Monam Agarwal <monamagarwal123@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Pull crypto updates from Herbert Xu:
"Here is the crypto update for 3.15:
- Added 3DES driver for OMAP4/AM43xx
- Added AVX2 acceleration for SHA
- Added hash-only AEAD algorithms in caam
- Removed tegra driver as it is not functioning and the hardware is
too slow
- Allow blkcipher walks over AEAD (needed for ARM)
- Fixed unprotected FPU/SSE access in ghash-clmulni-intel
- Fixed highmem crash in omap-sham
- Add (zero entropy) randomness when initialising hardware RNGs
- Fixed unaligned ahash comletion functions
- Added soft module depedency for crc32c for initrds that use crc32c"
* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (60 commits)
crypto: ghash-clmulni-intel - use C implementation for setkey()
crypto: x86/sha1 - reduce size of the AVX2 asm implementation
crypto: x86/sha1 - fix stack alignment of AVX2 variant
crypto: x86/sha1 - re-enable the AVX variant
crypto: sha - SHA1 transform x86_64 AVX2
crypto: crypto_wq - Fix late crypto work queue initialization
crypto: caam - add missing key_dma unmap
crypto: caam - add support for aead null encryption
crypto: testmgr - add aead null encryption test vectors
crypto: export NULL algorithms defines
crypto: caam - remove error propagation handling
crypto: hash - Simplify the ahash_finup implementation
crypto: hash - Pull out the functions to save/restore request
crypto: hash - Fix the pointer voodoo in unaligned ahash
crypto: caam - Fix first parameter to caam_init_rng
crypto: omap-sham - Map SG pages if they are HIGHMEM before accessing
crypto: caam - Dynamic memory allocation for caam_rng_ctx object
crypto: allow blkcipher walks over AEAD data
crypto: remove direct blkcipher_walk dependency on transform
hwrng: add randomness to system from rng sources
...
|
|
(struct caam_ctx) ctx->key_dma needs to be unmapped
when context is cleaned up.
Signed-off-by: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Add support for the following combinations:
-encryption: null
-authentication: md5, sha* (1, 224, 256, 384, 512)
Signed-off-by: Tudor Ambarus <tudor.ambarus@freescale.com>
Signed-off-by: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Commit 61bb86bba169507a5f223b94b9176c32c84b4721
("crypto: caam - set descriptor sharing type to SERIAL")
changed the descriptor sharing mode from SHARE_WAIT to SHARE_SERIAL.
All descriptor commands that handle the "ok to share" and
"error propagation" settings should also go away, since they have no
meaning for SHARE_SERIAL.
Signed-off-by: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Found by the kbuild test robot, the first argument to caam_init_rng
has a spurious ampersand.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
HIGHMEM pages may not be mapped so we must kmap them before accessing.
This resolves a random OOPs error that was showing up during OpenSSL SHA tests.
Signed-off-by: Joel Fernandes <joelf@ti.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
This patch allocates memory from DMAable region to the caam_rng_ctx object,
earlier it had been statically allocated which resulted in errorneous
behaviour on inserting the caamrng module at the runtime.
Signed-off-by: Nitesh Lal <NiteshNarayanLal@freescale.com>
Acked-by: Ruchika Gupta <ruchika.gupta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Optimize the hashing operation in the MXS-DCP by doing two adjustments:
1) Given that the output buffer for the hash is now always correctly aligned,
we can just use the buffer for the DCP DMA to store the resulting hash.
We thus get rid of one copying of data. Moreover, we remove an entry from
dcp_coherent_block{} and thus lower the memory footprint of the driver.
2) We map the output buffer for the hash for DMA only in case we will output
the hash, not always, as it was now.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: David S. Miller <davem@davemloft.net>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Shawn Guo <shawn.guo@linaro.org>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
The DCP needs the bounce buffers, DMA descriptors and result buffers aligned
to 64 bytes (yet another hardware limitation). Make sure they are aligned by
properly aligning the structure which contains them during allocation.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: David S. Miller <davem@davemloft.net>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Shawn Guo <shawn.guo@linaro.org>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Make omap_des_copy_needed(), omap_des_copy_sgs(), because these
functions are used only in this file.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Joel Fernandes <joelf@ti.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Use SIMPLE_DEV_PM_OPS macro in order to make the code simpler.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Use SIMPLE_DEV_PM_OPS macro in order to make the code simpler.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Use SIMPLE_DEV_PM_OPS macro in order to make the code simpler.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
coherent memory
1) SSYNC instruction is blackfin specific and takes no effect in this driver.
2) DMA descriptor and SG middle buffer are in DMA coherent memory. No need
to flush.
3) Turn kzalloc, ioremap and request_irq into managed device APIs respectively.
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
When the crypto layer is able to queue up a command for processing
by the CCP on the initial call to ccp_crypto_enqueue_request and
the CCP returns -EBUSY, then if the backlog flag is not set the
command needs to be freed and not added to the active command list.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
Invoke the callback routine associated with the crypto context
if an error is encountered sending the command to the CCP during
backlog processing. This is needed to free any resources used
by the command.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
|
If a CCP command has been queued for processing at the
crypto layer then, when dequeueing it for processing, the
"can backlog" flag must be set so that the request isn't
lost if the CCP backlog queue limit is reached.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
|
