> > > > diff --git a/arch/x86/kernel/cpu/sgx/driver.c > > > b/arch/x86/kernel/cpu/sgx/driver.c > > > > index 7f8d1e11dbee..b5ffe104af4c 100644 > > > > --- a/arch/x86/kernel/cpu/sgx/driver.c > > > > +++ b/arch/x86/kernel/cpu/sgx/driver.c > > > > @@ -19,6 +19,7 @@ static int sgx_open(struct inode *inode, struct file > *file) > > > > struct sgx_encl *encl; > > > > int ret; > > > > > > > > + sgx_inc_usage_count(); > > > > > > This should be done at the end of sgx_open() where the open cannot fail, > > > since > > > sgx_release() is not called if sgx_open() failed AFAICT. > > > > > > Could you please elaborate a bit more on this? > > In case sgx_inc_usage_count fails, we dont allocate resources yet, so what > > is > > wrong? > > > > I haven't looked into (details of) the last patch yet, but for _this_ patch > only, doing sgx_inc_usage_count() at the beginning of sgx_open() will result > in > the usage count being increased even when sgx_open() fails at a later time. > Since sgx_release() is not called when sgx_open() fails, the usage count will > not be decreased correctly. > > No?
Yes, I already yesterday understood that you meant *this* patch and yes, you are correct in this patch it must be moved to the end that we don't increment in case of failures.
