github-jwinterm/xmr-stak
1
0
Fork 0
mirror of https://github.com/jwinterm/xmr-stak.git synced 2026-10-17 05:41:41 +02:00
Code Issues Packages Projects Releases Wiki Activity

avoid possible illegal memory access

Browse Source 
Is is possible that the number of results are greater than the result output array,
in this case invalid memory can be access within the device and on the host side.
This commit is contained in:
psychocrypt
2017-10-06 21:50:51 +02:00
parent 83752eccc0
commit 611378eb96
2 changed files with 31 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
xmrstak/backend/amd/amd_gpu/gpu.cpp
+4
View File
@@ -912,6 +912,10 @@ size_t XMRRunJob(GpuContext* ctx, cl_uint* HashOutput)
}
clFinish(ctx->CommandQueues);
auto & numHashValues = HashOutput[0xFF];
// avoid out of memory read, we have only storage for 0xFF results
if(numHashValues > 0xFF)
numHashValues = 0xFF;
ctx->Nonce += g_intensity;
return ERR_SUCCESS;
xmrstak/backend/amd/amd_gpu/opencl/cryptonight.cl
+27 -8
View File
@@ -552,7 +552,7 @@ __kernel void cn2(__global uint4 *Scratchpad, __global ulong *states, __global u
states += 25 * gIdx;
Scratchpad += gIdx * (0x80000 >> 2);
for(int i = get_local_id(0); i < 256; i += WORKSIZE)
{
const uint tmp = AES0_C[i];
@@ -562,7 +562,7 @@ __kernel void cn2(__global uint4 *Scratchpad, __global ulong *states, __global u
AES3[i] = rotate(tmp, 24U);
}
barrier(CLK_LOCAL_MEM_FENCE);
#if defined(__Tahiti__) || defined(__Pitcairn__)
for(int i = 0; i < 4; ++i) ((ulong *)ExpandedKey2)[i] = states[i + 4];
@@ -677,8 +677,12 @@ __kernel void Skein(__global ulong *states, __global uint *BranchBuf, __global u
//vstore8(p, 0, output);
if(as_uint16(p).s7 <= Target) output[atomic_inc(output + 0xFF)] = BranchBuf[idx] + get_global_offset(0);
if(as_uint16(p).s7 <= Target)
{
ulong outIdx = atomic_inc(output + 0xFF);
if(outIdx < 0xFF)
output[outIdx] = BranchBuf[idx] + get_global_offset(0);
}
mem_fence(CLK_GLOBAL_MEM_FENCE);
}
@@ -743,7 +747,12 @@ __kernel void JH(__global ulong *states, __global uint *BranchBuf, __global uint
//output[2] = h7h;
//output[3] = h7l;
if(as_uint2(h7l).s1 <= Target) output[atomic_inc(output + 0xFF)] = BranchBuf[idx] + get_global_offset(0);
if(as_uint2(h7l).s1 <= Target)
{
ulong outIdx = atomic_inc(output + 0xFF);
if(outIdx < 0xFF)
output[outIdx] = BranchBuf[idx] + get_global_offset(0);
}
}
#define SWAP4(x) as_uint(as_uchar4(x).s3210)
@@ -811,7 +820,12 @@ __kernel void Blake(__global ulong *states, __global uint *BranchBuf, __global u
for(int i = 0; i < 8; ++i) h[i] = SWAP4(h[i]);
//for(int i = 0; i < 4; ++i) output[i] = ((ulong *)h)[i];
if(h[7] <= Target) output[atomic_inc(output + 0xFF)] = BranchBuf[idx] + get_global_offset(0);
if(h[7] <= Target)
{
ulong outIdx = atomic_inc(output + 0xFF);
if(outIdx < 0xFF)
output[outIdx] = BranchBuf[idx] + get_global_offset(0);
}
}
__kernel void Groestl(__global ulong *states, __global uint *BranchBuf, __global uint *output, uint Target, ulong Threads)
@@ -863,7 +877,12 @@ __kernel void Groestl(__global ulong *states, __global uint *BranchBuf, __global
for(int i = 0; i < 8; ++i) State[i] ^= tmp[i];
//for(int i = 0; i < 4; ++i) output[i] = State[i + 4];
if(as_uint2(State[7]).s1 <= Target) output[atomic_inc(output + 0xFF)] = BranchBuf[idx] + get_global_offset(0);
if(as_uint2(State[7]).s1 <= Target)
{
ulong outIdx = atomic_inc(output + 0xFF);
if(outIdx < 0xFF)
output[outIdx] = BranchBuf[idx] + get_global_offset(0);
}
}
)==="
)==="