Skip to content

C
clone-Linux

gsplhtlxg / clone-Linux

通知 2
Star 0
Fork 1
C
clone-Linux

体验新版 GitCode，发现更多精彩内容 >>

切换分支/标签
  1. 30 4月, 2007 14 次提交
    • S
      ieee1394: replace vmalloc by kmalloc in csr1212 · fd2f3bdd
      由 Stefan Richter 提交于 
      The biggest chunk ever allocated by CSR1212_MALLOC is 1024 Bytes +
sizeof(struct csr1212_csr_rom_cache) big.  Most of the time much
smaller data structures are allocated.  Therefore vmalloc is a waste.

The one exception is csr1212_append_new_cache() which is called to
append a chunk of CSR1212_EXTENDED_ROM_SIZE + sizeof(struct
csr1212_csr_rom_cache) if the currently allocated ROM cache is too
small.  CSR1212_EXTENDED_ROM_SIZE is generously defined as 256 kBytes.
In SVN commit 1220, Steve Kinneberg lowered this to 2 kBytes in the
config_rom_2.4 branch.  This same commit also switched CSR1212_MALLOC
from kmalloc to vmalloc in the SVN trunk branch:

> r1220 | kberg | 2004-05-31 01:51:44 +0200 (Mon, 31 May 2004) | 13 lines
>
> CSR1212 Extended ROM bug fixes:
> trunk line changes:
>   - Use vmalloc instead of kmalloc
>   - Change delayed_reset_bus() to operate in a work_queue instead of a
>     timer interrupt.
>   - Fix hpsb_allocate_and_register_addrspace() to not allocate space
>     on top of already allocated space.
>   - Fix problems in csr1212.c filling ConfigROM images when extend
>     ROMs are present.
> config-rom-2.4 changes:
>   - Changed extended rom allocation from 256K to 8K.
(It was actually 2 kB, not 8 kB.)
>   - Fix hpsb_allocate_and_register_addrspace() to not allocate space
>     on top of already allocated space.
>   - Fix problems in csr1212.c filling ConfigROM images when extend
>     ROMs are present.

I am now setting CSR1212_EXTENDED_ROM_SIZE to 2 kB minus the overhead of
struct csr1212_csr_rom_cache.  Note, this code path is not used by the
in-kernel drivers though.  raw1394 could trigger it, but the respective
libraw1394 functions don't exist yet.

Furthermore, userspace programs can replace the entire local ROM via
raw1394.  If kmalloc does not fulfill their needs --- well, tough luck.
I decree that nobody needs such huge extended ROMs.  (Extended ROMs are
defined by IEEE 1212 clause 7.7.18.  The spec does not impose
practically relevant restrictions on the size of extended ROM chunks.)

Another potentially demanding use of CSR1212_MALLOC is if external
FireWire devices come with Extended ROM entries.  If they are too big
for kmalloc (or have been too big for vmalloc) we just fail to read
their ROM.  This is quite unlikely though, to my knowledge.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      fd2f3bdd
    • S
      ieee1394: de-inline some functions · c1a37f2c
      由 Stefan Richter 提交于 
      This small reorganization of public csr1212 functions saves one
exported symbol and a few bytes in the driver modules.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      c1a37f2c
    • S
      ieee1394: stricter error checks in csr1212 · 64ff7123
      由 Stefan Richter 提交于 
      return -EINVAL becomes BUG_ON in checks of function call parameters.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      64ff7123
    • S
      ieee1394: csr1212: rename some types · 982610bd
      由 Stefan Richter 提交于 
      Use u8, u32 etc. instead of u_int8_t, csr1212_quad_t etc.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      982610bd
    • S
      ieee1394: drop csr1212's support for external compilation · 7fb9addb
      由 Stefan Richter 提交于 
      csr1212 was written to be compiled either as part of the ieee1394 kernel
driver or of an anticipated IEEE 1212 userspace library.  We now drop
support for the latter.  The costs in terms of code footprint and depth
of abstraction are not countered by any actual benefit.

Also remove some obsolete #includes.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      7fb9addb
    • S
      ieee1394: remove unused csr1212 code · 6c88e475
      由 Stefan Richter 提交于 
      Delete unused code.
Make some extern functions static.
Remove superfluous inline keywords.
Move private definitions from csr1212.h to csr1212.c.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      6c88e475
    • S
      ieee1394: small header cleanup · e167c88e
      由 Stefan Richter 提交于 
      Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      e167c88e
    • S
      ieee1394: move some comments from declaration to definition · afd6546d
      由 Stefan Richter 提交于 
      Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      afd6546d
    • S
      ieee1394: remove declarations of nonexisting functions · ef815334
      由 Stefan Richter 提交于 
      Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      ef815334
    • A
      ieee1394: sbp2: include fixes · f84c922b
      由 Andrew Morton 提交于 
      drivers/ieee1394/sbp2.c: In function 'sbp2util_access_timeout':
drivers/ieee1394/sbp2.c:399: error: 'TASK_INTERRUPTIBLE' undeclared (first use in this function)
drivers/ieee1394/sbp2.c:399: error: (Each undeclared identifier is reported only once
drivers/ieee1394/sbp2.c:399: error: for each function it appears in.)
drivers/ieee1394/sbp2.c:399: warning: implicit declaration of function 'signal_pending'
drivers/ieee1394/sbp2.c:399: warning: implicit declaration of function 'schedule_timeout'
drivers/ieee1394/sbp2.c: In function 'sbp2_prep_command_orb_sg':
drivers/ieee1394/sbp2.c:1438: warning: implicit declaration of function 'page_address'
drivers/ieee1394/sbp2.c:1438: warning: passing argument 2 of 'dma_map_single' makes pointer from integer without a cast
drivers/ieee1394/sbp2.c: In function 'sbp2_handle_status_write':
drivers/ieee1394/sbp2.c:1842: error: 'TASK_INTERRUPTIBLE' undeclared (first use in this function)

Possibly due to changes in -mm, but this file should explicitly include the
headers for the stuff it uses.
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> (brought into alphabetic order)
      f84c922b
    • S
      ieee1394: sbp2: move some memory allocations into non-atomic context · 3d269cb5
      由 Stefan Richter 提交于 
      When the command ORB pool is created, the ORB list won't be accessed
concurrently.  Therefore we don't have to take the spinlock there.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      3d269cb5
    • S
      ieee1394: sbp2: optimize DMA direction of s/g tables · 2446a79f
      由 Stefan Richter 提交于 
      Unlike the name suggests, "cmd->scatter_gather_element" holds only the
s/g table, not the actual s/g elements.  Since the table is only read
but never written by the device, DMA_BIDIRECTIONAL can be replaced by
DMA_TO_DEVICE which may be cheaper on some architectures.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      2446a79f
    • S
      ieee1394: sbp2: enforce 32bit DMA mapping · 0555659d
      由 Stefan Richter 提交于 
      In order to use OHCI-1394 physical DMA, all s/g elements, s/g tables,
ORBs, and response buffers have to reside within the first 4 GB of the
FireWire controller's physical address space.  Set the correct mask for
DMA mappings.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      0555659d
    • S
      ieee1394: sbp2: remove unnecessary alignments of struct members · f412bf44
      由 Stefan Richter 提交于 
      The members "dma_addr_t command_orb_dma" and "dma_addr_t sge_dma" of
sbp2.h::sbp2_command_info do not have to be aligned themselves --- only
the memory which they point to has to be.

The member "struct sbp2_command_orb command_orb" has to be aligned on
4 bytes boundary which is guaranteed because it contains u32 members.

The member "struct sbp2_unrestricted_page_table scatter_gather_element",
i.e. the SBP-2 s/g table, has to be aligned on 8 bytes boundary
according to the SBP-2 spec.  This is not a requirement for FireWire
controllers but could be expected by SBP-2 targets.

I see no need to align the members command_orb and
scatter_gather_element on CPU cacheline boundaries.  It could have
performance benefits, but on the other hand sbp2 has a somewhat wasteful
allocation scheme which should be optimized first before further tweaks
like cacheline alignments.  (E.g. don't always allocate SG_ALL s/g table
elements.)

Note, before as well as after the patch, the code relies on the
assumption that memory alignment in the virtual address space is
preserved in the physical address space after DMA mapping.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
      f412bf44
  3. 28 4月, 2007 26 次提交