Instead of shm_open.
Created attachment 350101 [details] Patch
Comment on attachment 350101 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=350101&action=review In this case, I would check for sys/memfd.h instead of linux/memfd.h and use memfd_create() directly if it's available (glibc 2.27 or newer) rather than using syscall(). Then you don't have to check for EINTR or ENOSYS. This way works too, of course. > Source/WebKit/Platform/unix/SharedMemoryUnix.cpp:128 > + if (errno == ENOSYS) > + return fileDescriptor; Surely this isn't right? You should just remove this condition and fall through to isMemFdAvailable = false, right? > Source/WebKit/Platform/unix/SharedMemoryUnix.cpp:154 > - WTFLogAlways("Failed to create shared memory file %s: %s", tempName.data(), strerror(errno)); > + WTFLogAlways("Failed to create shared memory file: %s", strerror(errno)); Drop the word "file"
(In reply to Michael Catanzaro from comment #2) > Comment on attachment 350101 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=350101&action=review > > In this case, I would check for sys/memfd.h instead of linux/memfd.h and use > memfd_create() directly if it's available (glibc 2.27 or newer) rather than > using syscall(). Then you don't have to check for EINTR or ENOSYS. > > This way works too, of course. I'm assuming this is available in more systems, no? > > Source/WebKit/Platform/unix/SharedMemoryUnix.cpp:128 > > + if (errno == ENOSYS) > > + return fileDescriptor; > > Surely this isn't right? You should just remove this condition and fall > through to isMemFdAvailable = false, right? Yes, I changed this several times and ended up with the wrong check. Good catch! > > Source/WebKit/Platform/unix/SharedMemoryUnix.cpp:154 > > - WTFLogAlways("Failed to create shared memory file %s: %s", tempName.data(), strerror(errno)); > > + WTFLogAlways("Failed to create shared memory file: %s", strerror(errno)); > > Drop the word "file" Sure.
(In reply to Carlos Garcia Campos from comment #3) > I'm assuming this is available in more systems, no? The syscall was added years before the glibc interface, but so what? It's nicer to use the glibc function, and there is a fallback, after all. syscall() is used in the sandbox code because it's really needed there, but here it's not: there is a fallback.
(In reply to Michael Catanzaro from comment #4) > (In reply to Carlos Garcia Campos from comment #3) > > I'm assuming this is available in more systems, no? > > The syscall was added years before the glibc interface, but so what? It's > nicer to use the glibc function, and there is a fallback, after all. > syscall() is used in the sandbox code because it's really needed there, but > here it's not: there is a fallback. I still don't understand why sandbox code can't use SharedMemory.
Committed r237922: <https://trac.webkit.org/changeset/237922>
<rdar://problem/45873944>