Port libcutils memset16/32 assembly SSE2 optimizations to x86_64
architecture. Ensures the same performance on 64-bit arch.
Change-Id: I874a71a884c0d28a152933ddff9cb886c9a6e99e
Signed-off-by: Henrik Smiding <henrik.smiding@intel.com>
Not seeding this each call should help reduce collisions when multiple
threads are calling ashmem_create_region. Also cleaned up code by
deleting gotos, and making formatting consistent.
Bug: 15394258
Change-Id: Iafdaea57b2317e0eb7c40d7b5595c523814fd88c
In some condition, some of FS would set disable_roll_forward option.
In this case, when we do power off or reboot, the power down progress
would stuck at android_reboot for 360 sec.
The reason is, remount_ro_done function check key word "rw" to determine
whether all the read-write FS has been remounted as read-only FS.
Unfortunately, it takes disable_roll_forward as a rw flag too by
mistake, so the check would always fail, until reach 360 sec threshold.
The patch would match "rw," from mount options, this can avoid the issue
successfully.
Change-Id: I409f9dc88c9ee5cf049615ea9dcccaf195b6bc5a
Signed-off-by: Hong-Mei Li <a21834@motorola.com>
After a disconnect, the initial blocking connect takes
a long time to return, while subsequent calls return
quicks. Switch to a non-blocking connect to make the
re-connect time more consistent and faster overall.
Change-Id: I21d02b22a8eb9a457c2f1fa95eb17894d5612ccd
Signed-off-by: Ken Lierman <ken.lierman@windriver.com>
Reviewed-by: Gumbel, Matthew K <matthew.k.gumbel@intel.com>
Reviewed-by: Jovanovic, Radivoje <radivoje.jovanovic@intel.com>
Reviewed-by: Boie, Andrew P <andrew.p.boie@intel.com>
Bionic needs to re-raise various signals, which means the si_code
debuggerd sees has been clobbered. If bionic sends us the original
si_code value, we can use that instead of the one we see when the
ptrace the crashed process' siginfo.
Change-Id: If116a6bc667d55a6fb39b74f96673292af4e4c8c
Adds Silvermont specific cache sizes for memset16/32 SSE optimization.
Change-Id: Ib5ea086d57544e74ac384ee1ef516b8511392f70
Signed-off-by: Henrik Smiding <henrik.smiding@intel.com>
A normal sequence of calls is as follows:
str_parms_create_str, str_parms_add_str, str_parms_destroy.
In some cases the destroy caused double free.
str_parms_add_str will clone the input and send it to hashmapPut
for storage. If hashmapPut did not store the strings it will raise
errno = ENOMEM and leave caller with ownership of the strings.
In any of these cases it will be safe to destroy the str_parms.
But what if it wasn't hashmapPut that said NOMEM? What if there
was a stale NOMEM already before a successful hashmapPut?
In that case the strings will be successfully added to the list
(if new), but when str_parms_add_str sees the NOMEM it will free
them anyway, leaving dangling pointers in the str_parms!!
It is the responsibility of the caller to clear errno before any
interesting call. This patch makes sure that str_parms_add_str
reacts only on errno emmitted from hashmapPut.
Change-Id: If87e4bcc482f09e1c66133d33517b152ebdac65f
We have benchmarking for memset in bionic itself, and a benchmarking
framework for other native benchmarking.
Change-Id: I241a288a96d3b2d37f3d51b04a519ac23f18a989
Eliminates various warnings from SELinux-related code.
Bug: 12587913
Change-Id: I28921f0ebd934324436609540d95ccef58552b64
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
str_parms_get_float did not return the output into
*val. Only output if returning with no error. Audit
shows no internal users of this library function
(cherry picked from commit 8e71ddeac0)
Change-Id: I14a3f08a098072a159dd93f85ead36b3f445816f
The kernel headers are already in the include path, and manually
adding them again will break on a multiarch build, where the
kernel headers may be different for each arch.
Change-Id: Ib13d96543f2c97b1b487a46e1748ceb2c5724169
When initiating a manual dump, the abort_msg_address parameter is random
garbage. Initialize this value to 0.
Change-Id: I9ba817ee6104f47cfcb2b746bfa6f02a259c86d7
__reboot is hidden for 64-bits and should be hidden on others
Change-Id: Ib6435269aac7fa726e49cc02f8f76d2be2875bc6
Signed-off-by: Pavel Chupin <pavel.v.chupin@intel.com>
Originally, if TARGET_ARCH_VARIANT is x86-atom, the SSE2 enhanced
memset is used. This patch extends this to all x86 processors which
support SSE2 (i.e. ARCH_X86_HAVE_SSE2 is true).
Indentation added to the ifeq cases to make this easier to read.
Change-Id: I05f49e237a95359d3f2e3216b037e3fc1a0fbcb0
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Before this change, FUSE lookup() would have the side effect of
creating the directory on behalf of apps. This resulted in most
directories being created just by Settings trying to measure disk
space. Instead, we're switching to have vold do directory creation
when an app doesn't have enough permissions.
Create fs_mkdirs() utility to create all parent directories in a
path as needed. Allow traversal (+x) into /storage directories.
Fix FUSE derived permissions to be case insensitive. Mark well-known
directories as .nomedia when created.
Bug: 10577808, 10330221
Change-Id: I53114f2e63ffbe6de4ba6a72d94a232523231cad
Henrik Smiding