| Message ID | 914642975.1359462.1466295321636.JavaMail.zimbra@raptorengineeringinc.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 19 June 2016 at 01:15, Timothy Pearson <tpearson@raptorengineering.com> wrote: > This matches the calling conventions in the Linux kernel and > resolves select() hangs on i386/x86_64 guests. > > Signed-off-by: Timothy Pearson <tpearson@raptorengineering.com> > --- > linux-user/syscall.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index 2968b57..644a490 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -4785,6 +4785,7 @@ static abi_long do_ioctl(int fd, int cmd, abi_long arg) > void *argptr; > > ie = ioctl_entries; > + > for(;;) { > if (ie->target_cmd == 0) { > gemu_log("Unsupported ioctl: cmd=0x%04lx\n", (long)cmd); > @@ -7899,7 +7900,7 @@ abi_long do_syscall(void *cpu_env, int num, abi_long arg1, > break; > #if defined(TARGET_NR_select) > case TARGET_NR_select: > -#if defined(TARGET_S390X) || defined(TARGET_ALPHA) > +#if defined(TARGET_S390X) || defined(TARGET_ALPHA) || defined(TARGET_I386) || defined(TARGET_X86_64) > ret = do_select(arg1, arg2, arg3, arg4, arg5); > #else > { I think we can fix this in a cleaner way. This is actually the result of there being two select syscalls: * the one which takes a struct with the params, which the kernel calls 'old_select' and which we seem to have decided to call TARGET_NR_select * the one which takes the arguments separately, which the kernel calls 'select' and which we seem to use TARGET_NR__newselect for So I think we should fix this by: (1) making sure all archs (including x86, alpha and s390) define TARGET_NR_select and TARGET_NR__newselect correctly (2) removing this ifdef entirely (so TARGET_NR_select is always "use the struct" for any arch that defines it, and TARGET_NR__newselect is always "direct args"), and similarly for the equivalent ifdef in linux-user/syscall.c I think we should also change our naming to TARGET_NR_oldselect and TARGET_NR_select to match the kernel. thanks -- PMM
diff --git a/linux-user/syscall.c b/linux-user/syscall.c index 2968b57..644a490 100644 --- a/linux-user/syscall.c +++ b/linux-user/syscall.c @@ -4785,6 +4785,7 @@ static abi_long do_ioctl(int fd, int cmd, abi_long arg) void *argptr; ie = ioctl_entries; + for(;;) { if (ie->target_cmd == 0) { gemu_log("Unsupported ioctl: cmd=0x%04lx\n", (long)cmd); @@ -7899,7 +7900,7 @@ abi_long do_syscall(void *cpu_env, int num, abi_long arg1, break; #if defined(TARGET_NR_select) case TARGET_NR_select: -#if defined(TARGET_S390X) || defined(TARGET_ALPHA) +#if defined(TARGET_S390X) || defined(TARGET_ALPHA) || defined(TARGET_I386) || defined(TARGET_X86_64) ret = do_select(arg1, arg2, arg3, arg4, arg5); #else {
This matches the calling conventions in the Linux kernel and resolves select() hangs on i386/x86_64 guests. Signed-off-by: Timothy Pearson <tpearson@raptorengineering.com> --- linux-user/syscall.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)